The rapid development of deep learning has significantly improved salient object detection (SOD) combining both RGB and thermal (RGB-T) images. However, existing Transformer-based RGB-T SOD models with quadratic complexity are memory-intensive, limiting their application in high-resolution bimodal feature fusion. To overcome this limitation, we propose a purely Fourier Transform-based model, namely Deep Fourier-embedded Network (FreqSal), for accurate RGB-T SOD. Specifically, we leverage the efficiency of Fast Fourier Transform with linear complexity to design three key components: (1) To fuse RGB and thermal modalities, we propose Modal-coordinated Perception Attention, which aligns and enhances bimodal Fourier representation in multiple dimensions; (2) To clarify object edges and suppress noise, we design Frequency-decomposed Edge-aware Block, which deeply decomposes and filters Fourier components of low-level features; (3) To accurately decode features, we propose Fourier Residual Channel Attention Block, which prioritizes high-frequency information while aligning channel-wise global relationships. Additionally, even when converged, existing deep learning-based SOD models’ predictions still exhibit frequency gaps relative to ground-truth. To address this problem, we propose Co-focus Frequency Loss, which dynamically weights hard frequencies during edge frequency reconstruction by cross-referencing bimodal edge information in the Fourier domain. Extensive experiments on ten bimodal SOD benchmark datasets demonstrate that FreqSal outperforms twenty-nine existing state-of-the-art bimodal SOD models. Comprehensive ablation studies further validate the value and effectiveness of our newly proposed components. The code is available at https://github.com/JoshuaLPF/FreqSal.

Semi-supervised segmentation is gaining popularity in medical image analysis due to challenges in data acquisition and annotation. However, most methods focus on generating additional training pairs from unlabeled data through augmentation or perturbation for contrastive learning, often overlooking the unique characteristics and inherent priors of medical images. We identified two key anatomical priors in fetal cardiac ultrasound images: (1) anatomies have consistent shapes and locations due to standard views captured by sonographers from fixed angles; (2) category pixels are densely clustered, with each structure appearing only once per image. We propose AP-Net, which uses an anatomical prior generation module, a prior-feature fusion module, and a category-aware cropping strategy to effectively leverage these anatomical priors. Experiments on a real-world fetal cardiac ultrasound dataset show that AP-Net outperforms classical supervised and leading semi-supervised methods, with each component enhancing its performance.

Although brain imaging has provided crucial insights into Alzheimer’s disease (AD) progression, its limitations in capturing molecular changes have motivated the need to integrate and interpret transcriptomic data, which can reveal early-stage molecular disruptions. By interpreting both structural and gene expression patterns, there is potential to gain a more comprehensive understanding of the disease and identify interactions that drive AD pathogenesis. This study introduces a novel deep learning framework that integrates structural magnetic resonance imaging (sMRI) and gene expression (GE) data for AD prediction. The model extracts features from both data modalities and employs a novel approach to quantify cross-modal interactions. Our multimodal architecture achieves 83.6% accuracy in AD prediction, demonstrating its effectiveness in combining imaging and transcriptomic data. We present a technique for interpreting interactions between specific brain regions and gene expression patterns, providing insights into their joint contribution to AD risk. This approach enables the identification of both established and potential new AD biomarkers, spanning structural brain changes and transcriptomic alterations. By offering a deeper understanding of AD’s molecular and structural aspects, our work advances the field of multimodal biomarker discovery in neurodegenerative diseases and paves the way for more comprehensive early diagnosis strategies.

The overexpression of B-cell lymphoma-extra large (BCL-xL), an anti-apoptotic protein, plays a pivotal role in various cancers by inhibiting apoptosis and promoting tumor progression. We introduce Latent Diffusion Model (LDM) specifically designed to generate apoptosis-inducing Anticancer Peptides (ACPs) targeting BCL-xL. By applying a diffusion process to high-dimensional embeddings generated by ESM-2, a state-of-the-art Protein Language Model (PLM), our approach successfully generates peptides with optimal physicochemical properties, even when trained on a relatively small dataset. In comparison to Variational Autoencoder (VAE) and Wasserstein Autoencoder (WAE), our architecture not only mitigates the limitations associated with small datasets but also produces peptides that closely align with desired physiochemical properties. Evaluation against key metrics—molecular weight, isoelectric point, net charge at pH 7, grand average of hydropathy, instability index, and anticancer peptide score,demonstrates that our model consistently outperforms both VAE and WAE, offering a promising strategy for the development of novel anticancer therapies.

Federated learning (FL) enables multiple clients to collaboratively train machine learning models without exposing local data, balancing performance and privacy. However, domain shift and label heterogeneity across clients often hinder the generalization of the aggregated global model. Recently, large-scale vision-language models like CLIP have shown strong zero-shot classification capabilities, raising the question of how to effectively fine-tune CLIP across domains in a federated setting. In this work, we propose an adaptive federated prompt tuning framework, FedDEAP, to enhance CLIP's generalization in multi-domain scenarios. Our method includes the following three key components: (1) To mitigate the loss of domain-specific information caused by label-supervised tuning, we disentangle semantic and domain-specific features in images by using semantic and domain transformation networks with unbiased mappings; (2) To preserve domain-specific knowledge during global prompt aggregation, we introduce a dual-prompt design with a global semantic prompt and a local domain prompt to balance shared and personalized information; (3) To maximize the inclusion of semantic and domain information from images in the generated text features, we align textual and visual representations under the two learned transformations to preserve semantic and domain consistency. Theoretical analysis and extensive experiments on four datasets demonstrate the effectiveness of our method in enhancing the generalization of CLIP for federated image recognition across multiple domains.

Deep learning has significantly advanced the field of Speech Emotion Recognition (SER), yet its efficacy in cross-corpus scenarios remains a challenge. To overcome this limitation, recent studies demonstrate the success of multitask learning, which uses auxiliary tasks to reduce difference between source and target dataset (or transfer knowledge from source to target datasets). Despite the efforts, the overall accuracy for cross-corpus SER is still relatively low and needs attention. To improve performance, we propose a multitask framework with SER as the primary task and contrastive learning and information maximization as auxiliary tasks. We design the auxiliary tasks innovatively to use the target data without emotional labels to develop a better understanding of the target data. The core of our multitask framework is a pre-trained transformer. While transformers have gained attention in SER, their application to cross-corpus scenarios is still limited. Multimodal approaches for cross-corpus scenario is substantially limited as well. We use text as the second modality, developing separate multitask transformers for audio and text and conduct decision-level fusion during inference. We use publicly available and widely used speech corpora, including the IEMOCAP, MSP-IMPROV and EMO-DB databases. The results demonstrate the benefits of the proposed approach, achieving improved performance on the benchmark databases in cross-corpus settings.

Unmanned aerial vehicle (UAV)-based bi-modal salient object detection (BSOD) aims to segment salient objects in a scene utilizing complementary cues in unaligned RGB and thermal image pairs. However, the high computational expense of existing UAV-based BSOD models limits their applicability to real-world UAV devices. To address this problem, we propose an efficient Fourier filter network with contrastive learning that achieves both real-time and accurate performance. Specifically, we first design a semantic contrastive alignment loss to align the two modalities at the semantic level, which facilitates mutual refinement in a parameter-free way. Second, inspired by the fast Fourier transform that obtains global relevance in linear complexity, we propose synchronized alignment fusion, which aligns and fuses bi-modal features in the channel and spatial dimensions by a hierarchical filtering mechanism. Our proposed model, AlignSal, reduces the number of parameters by 70.0%, decreases the floating point operations by 49.4%, and increases the inference speed by 152.5% compared to the cutting-edge BSOD model (i.e., MROS). Extensive experiments on the UAV RGB-T 2400 and seven bi-modal dense prediction datasets demonstrate that AlignSal achieves both real-time inference speed and better performance and generalizability compared to nineteen state-of-the-art models across most evaluation metrics. In addition, our ablation studies further verify AlignSal’s potential in boosting the performance of existing aligned BSOD models on UAV-based unaligned data. The code is available at: https://github.com/JoshuaLPF/AlignSal.

Exploring complementary information between RGB and thermal/depth modalities is crucial for bi-modal salient object detection (BSOD). However, the distinct characteristics of different modalities often lead to large differences in information distributions. Existing models, which rely on convolutional operations or plug-and-play attention mechanisms, struggle to address this issue. To overcome this challenge, we rethink the relationship between information complementarity and long-range relevance, and propose a uniform broad-view Twins Transformer Network (TwinsTNet) for accurate BSOD. Specifically, to efficiently fuse bi-modal information, we first design the Cross-Modal Federated Attention (CMFA), which mines complementary cues across modalities through element-wise global dependency. Second, to ensure accurate modality fusion, we propose the Semantic Consistency Attention Loss, which supervises the co-attention feature in CMFA using the ground-truth-generated attention map. Additionally, existing BSOD models lack the exploration of inter-layer interactions, for which we propose the Cross-Scale Retracing Attention (CSRA), which retrieves query-relevant information from stacked features of all previous layers, enabling flexible cross-layer interactions. The cooperation between CMFA and CSRA mitigates inductive bias in both modality and layer dimensions, enhancing TwinsTNet's representational capability. Extensive experiments demonstrate that TwinsTNet outperforms twenty-two existing state-of-the-art models on ten BSOD benchmark datasets. The code is available at: https://github.com/JoshuaLPF/TwinsTNet.

The detection of cardiac phase in ultrasound videos, identifying end-systolic (ES) and end-diastolic (ED) frames, is a critical step in assessing cardiac function, monitoring structural changes, and diagnosing congenital heart disease. Current popular methods use recurrent neural networks to track dependencies over long sequences for cardiac phase detection, but often overlook the short-term motion of cardiac valves that sonographers rely on. In this paper, we propose a novel optical flow-enhanced Mamba U-net framework, designed to utilize both short-term motion and long-term dependencies to detect the cardiac phase in ultrasound videos. We utilize optical flow to capture the short-term motion of cardiac muscles and valves between adjacent frames, enhancing the input video. The Mamba layer is employed to track long-term dependencies across cardiac cycles. We then develop regression branches using the U-Net architecture, which integrates short-term and long-term information while extracting multi-scale features. Using this method, we can generate regression scores for each frame and identify keyframes (i.e., ES and ED frames). Additionally, we design a keyframe weighted loss function to guide the network to focus more on keyframes rather than intermediate period frames. Our method demonstrates superior performance compared to advanced baseline methods, achieving frame mismatches of 1.465 frames for ES and 0.842 frames for ED in the Fetal Echocardiogram dataset, where heart rates are higher and phase changes occur rapidly, and 2.444 frames and 2.072 frames in the publicly available adult Echonet-Dynamic dataset. Its accuracy and robustness in both fetal and adult datasets highlight its potential for clinical application.

Medical image segmentation plays a vital role in medical image analysis. However, it is impractical to build a large-scale centralized segmentation dataset due to the privacy of medical images. Federated learning (FL) aims to train a shared model of isolated clients without local data exchange which aligns well with the scarcity and privacy characteristics of medical images. Moreover, there is a large amount of unlabeled data in clients due to the difficulty in annotating medical images. Federated semi-supervised learning (FSSL) can leverage the unlabeled data of clients to improve the performance of the global model. Many existing FSSL methods apply the complicated semi-supervised learning protocols and some of them neglect the problem of data heterogeneity in FL. In this paper, we propose a novel federated semi-supervised learning framework for medical image segmentation incorporating intra-client and inter-client consistency learning. The intra-client consistency learning can introduce global data noise in data augmentation which can improve the generalization ability of the model and reduce the impact of data heterogeneity. The inter-client consistency learning is proposed to expand the feature search space and learn the ensemble knowledge of different clients. The two consistency learning mechanisms are achieved with the assistance of a Variational Autoencoder (VAE) trained collaboratively by clients. The experimental results illustrate that our method outperforms the state-of-the-art methods under different FSSL settings. The code is available at https://github.com/zyb98/FV2IC.

Heterogeneity is present in Alzheimer’s disease (AD), making it challenging to study. To address this, we propose a graph neural network (GNN) approach to identify disease subtypes from magnetic resonance imaging (MRI) and functional MRI (fMRI) scans. Subtypes are identified by encoding the patients’ scans in brain graphs (via cortical similarity networks) and clustering the representations learnt by the GNN. These subtyping information are used to construct population graphs for an ensemble of local networks, each producing intermediate predictions that are subsequently combined to produce the model’s final decision. Using MRI and fMRI scans from two datasets on AD, we demonstrate that our proposed architecture outperforms existing methods. Three subtypes of AD were identified and left cuneus was found to be a consistent class-wide biomarker. Subtype-specific biomarkers produced by our method further revealed deeper insights, including a unique subtype with significant degeneration in the left isthmus cingulate cortex.

De novo generation of molecules is a crucial task in drug discovery. The blossom of deep learning-based generative models, especially diffusion models, has brought forth promising advancements in de novo drug design by finding optimal molecules in a directed manner. However, due to the complexity of chemical space, existing approaches can only generate extremely small molecules. In this study, we propose a Graph Latent Diffusion Model (GLDM) that operates a diffusion model in the latent space modeled by a pretrained autoencoder. Applying diffusion processs on latent representations rather than original molecular graphs, GLDM improves training efficiency and enables generation of larger drug-like molecules. GLDM achieves state-of-the-art results on GuacaMol benchmarks.

Predicting fluid cognition via neuroimaging data is essential for understanding the neural mechanisms underlying various complex cognitions in the human brain. Both brain functional connectivity (FC) and structural connectivity (SC) provide distinct neural mechanisms for fluid cognition. In addition, interactions between SC and FC within distributed association regions are related to improvements in fluid cognition. However, existing learning-based methods that leverage both modality-specific embeddings and high-order interactions between the two modalities for prediction are scarce. To tackle these challenges, this study proposes an end-to-end brain structure-function interaction network that incorporates both modality-specific embeddings and structure-function interactions to predict fluid cognition. In this model, we generate embeddings from both FC and SC separately using a graph convolution encoder-decoder module. Subsequently, we learn the interactive weights between corresponding regions of FC and SC, reflecting the coupling strength, by employing an interactive module on the embeddings of both modalities. A novel graph structure - utilizing modality-specific embeddings and interactive weights - is constructed and used for the final prediction. Experimental results demonstrate that our proposed method outperforms other state-of-the-art methods employed on uni-modal and multi-modal brain features. We further identify that strong structure-function coupling in the inferior frontal, postcentral, superior temporal and cingulate cortices are associated with fluid intelligence.

Discovering hit molecules with desired biological activity in a directed manner is a promising but profound task in computer-aided drug discovery. Inspired by recent generative AI approaches, particularly Diffusion Models (DM), we propose Graph Latent Diffusion Model (GLDM)-a latent DM that preserves both the effectiveness of autoencoders of compressing complex chemical data and the DM's capabilities of generating novel molecules. Specifically, we first develop an autoencoder to encode the molecular data into low-dimensional latent representations and then train the DM on the latent space to generate molecules inducing targeted biological activity defined by gene expression profiles. Manipulating DM in the latent space rather than the input space avoids complicated operations to map molecule decomposition and reconstruction to diffusion processes, and thus improves training efficiency. Experiments show that GLDM not only achieves outstanding performances on molecular generation benchmarks, but also generates samples with optimal chemical properties and potentials to induce desired biological activity.

Congenital heart disease (CHD) is the most common congenital disability affecting healthy development and growth, even resulting in pregnancy termination or fetal death. Recently, deep learning techniques have made remarkable progress to assist in diagnosing CHD. One very popular method is directly classifying fetal ultrasound images, recognized as abnormal and normal, which tends to focus more on global features and neglects semantic knowledge of anatomical structures. The other approach is segmentation-based diagnosis, which requires a large number of pixel-level annotation masks for training. However, the detailed pixel-level segmentation annotation is costly or even unavailable. Based on the above analysis, we propose SKGC, a universal framework to identify normal or abnormal four-chamber heart (4CH) images, guided by a few annotation masks, while improving accuracy remarkably. SKGC consists of a semantic-level knowledge extraction module (SKEM), a multi-knowledge fusion module (MFM), and a classification module (CM). SKEM is responsible for obtaining high-level semantic knowledge, serving as an abstract representation of the anatomical structures that obstetricians focus on. MFM is a lightweight but efficient module that fuses semantic-level knowledge with the original specific knowledge in ultrasound images. CM classifies the fused knowledge and can be replaced by any advanced classifier. Moreover, we design a new loss function that enhances the constraint between the foreground and background predictions, improving the quality of the semantic-level knowledge. Experimental results on the collected real-world NA-4CH and the publicly FEST datasets show that SKGC achieves impressive performance with the best accuracy of 99.68% and 95.40%, respectively. Notably, the accuracy improves from 74.68% to 88.14% using only 10 labeled masks.

Multi-omics data provides a wealth of information concerning disease mechanisms, which benefits the exploration of the intricate molecular phenomena underlying diseases. In recent years, considerable endeavors have been directed towards the combination of graph convolutional network, which has the powerful ability to gather information, with multi-omics learning methods to obtain more reliable results. For achieving this pursuit, an essential challenge is data integration. Against this backdrop, we propose a unified framework named multi-level knowledge integration with graph convolutional network, which effectively incorporates multiple prior knowledge and omics data to learn an intrinsic representation. In specific, the model consists of two subnetworks: an attribute-level module and a sample-level module. The former firstly aggregates the knowledge given by the prior biological graphs into low-dimensional embeddings, and then maximizes the consistency between these prior views via optimizing a contrastive loss for attaining the attribute-based representations. The latter leverages an encoder to dimensionalize the original multi-omics data to attain more dominant sample knowledge, and subsequently utilizes another contrastive loss to align these representations between multiple omics for learning the global sample-level information. Comprehensive experiments are performed to show that the proposed model surpasses other state-of-the-art methods.

With the development of biotechnology, a large amount of multi-omics data have been collected for precision medicine. There exists multiple graph-based prior biological knowledge about omics data, such as gene-gene interaction networks. Recently, there has been an increasing interest in introducing graph neural networks (GNNs) into multi-omics learning. However, existing methods have not fully exploited these graphical priors since none have been able to integrate knowledge from multiple sources simultaneously. To solve this problem, we propose a multi-omics data analysis framework by incorporating multiple prior knowledge into graph neural network (MPK-GNN). To the best of our knowledge, this is the first attempt to introduce multiple prior graphs into multi-omics data analysis. Specifically, the proposed method contains four parts: (1) a feature-level learning module to aggregate information from prior graphs; (2) a projection module to maximize the agreement among prior networks by optimizing a contrastive loss; (3) a sample-level module to learn a global representation from input multi-omics features; (4) a task-specific module to flexibly extend MPK-GNN for various downstream multi-omics analysis tasks. Finally, we verify the effectiveness of the proposed multi-omics learning algorithm on the cancer molecular subtype classification task. Experimental results show that MPK-GNN outperforms other state-of-the-art algorithms, including multi-view learning methods and multi-omics integrative approaches.

In contrast to traditional adversarial learning (AL) which learns speaker-invariant representations, this paper proposes cascaded adversarial learning (CAL) which learns speaker-invariant emotion data for speaker independent emotion recognition (SIER) tasks. CAL is a dual cascaded network architecture where the output of the transformation network is fed as input to the classification network. Transformation network transforms original speech emotion to speaker-invariant emotion data by implementing an AL strategy with an encoder-decoder architecture. The classification network predicts the emotion from the speaker-invariant emotion data (output of the transformation network). We argue that the speaker-invariant emotion data realized by transformation network has less variation than the original speech emotion data and therefore are conducive for SIER as it improve generalization capability. To our knowledge this is the first time a dual cascaded network has been used for SIER and demonstrate state-of-the-art performances for SIER on Emo-DB and RAVDESS datasets.

MOTIVATION: While it has been well established that drugs affect and help patients differently, personalized drug response predictions remain challenging. Solutions based on single omics measurements have been proposed, and networks provide means to incorporate molecular interactions into reasoning. However, how to integrate the wealth of information contained in multiple omics layers still poses a complex problem. RESULTS: We present DrDimont, Drug response prediction from Differential analysis of multi-omics networks. It allows for comparative conclusions between two conditions and translates them into differential drug response predictions. DrDimont focuses on molecular interactions. It establishes condition-specific networks from correlation within an omics layer that are then reduced and combined into heterogeneous, multi-omics molecular networks. A novel semi-local, path-based integration step ensures integrative conclusions. Differential predictions are derived from comparing the condition-specific integrated networks. DrDimont's predictions are explainable, i.e. molecular differences that are the source of high differential drug scores can be retrieved. We predict differential drug response in breast cancer using transcriptomics, proteomics, phosphosite and metabolomics measurements and contrast estrogen receptor positive and receptor negative patients. DrDimont performs better than drug prediction based on differential protein expression or PageRank when evaluating it on ground truth data from cancer cell lines. We find proteomic and phosphosite layers to carry most information for distinguishing drug response. AVAILABILITY AND IMPLEMENTATION: DrDimont is available on CRAN: https://cran.r-project.org/package=DrDimont. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Cancers are genetically heterogeneous, so anticancer drugs show varying degrees of effectiveness on patients due to their differing genetic profiles. Knowing patient's responses to numerous cancer drugs are needed for personalized treatment for cancer. By using molecular profiles of cancer cell lines available from Cancer Cell Line Encyclopedia (CCLE) and anticancer drug responses available in the Genomics of Drug Sensitivity in Cancer (GDSC), we will build computational models to predict anticancer drug responses from molecular features. RESULTS: We propose a novel deep neural network model that integrates multi-omics data available as gene expressions, copy number variations, gene mutations, reverse phase protein array expressions, and metabolomics expressions, in order to predict cellular responses to known anti-cancer drugs. We employ a novel graph embedding layer that incorporates interactome data as prior information for prediction. Moreover, we propose a novel attention layer that effectively combines different omics features, taking their interactions into account. The network outperformed feedforward neural networks and reported 0.90 for [Formula: see text] values for prediction of drug responses from cancer cell lines data available in CCLE and GDSC. CONCLUSION: The outstanding results of our experiments demonstrate that the proposed method is capable of capturing the interactions of genes and proteins, and integrating multi-omics features effectively. Furthermore, both the results of ablation studies and the investigations of the attention layer imply that gene mutation has a greater influence on the prediction of drug responses than other omics data types. Therefore, we conclude that our approach can not only predict the anti-cancer drug response precisely but also provides insights into reaction mechanisms of cancer cell lines and drugs as well.

BACKGROUND: Protein-protein interaction (PPI) prediction is an important task towards the understanding of many bioinformatics functions and applications, such as predicting protein functions, gene-disease associations and disease-drug associations. However, many previous PPI prediction researches do not consider missing and spurious interactions inherent in PPI networks. To address these two issues, we define two corresponding tasks, namely missing PPI prediction and spurious PPI prediction, and propose a method that employs graph embeddings that learn vector representations from constructed Gene Ontology Annotation (GOA) graphs and then use embedded vectors to achieve the two tasks. Our method leverages on information from both term-term relations among GO terms and term-protein annotations between GO terms and proteins, and preserves properties of both local and global structural information of the GO annotation graph. RESULTS: We compare our method with those methods that are based on information content (IC) and one method that is based on word embeddings, with experiments on three PPI datasets from STRING database. Experimental results demonstrate that our method is more effective than those compared methods. CONCLUSION: Our experimental results demonstrate the effectiveness of using graph embeddings to learn vector representations from undirected GOA graphs for our defined missing and spurious PPI tasks.

Protein-protein interaction (PPI) prediction is a key step towards many bioinformatics applications including prediction of protein functions and drug-disease interactions. However, previous research on PPI prediction rarely considered missing and spurious interactions in PPI networks. To address these two issues, we define two corresponding tasks, namely missing PPI prediction and spurious PPI prediction, and propose a novel method that employs graph embeddings to learn vector representations from constructed Gene Ontology (GO) annotation graphs. Our method leverages the information from both term-term relations among GO terms and term-protein annotations between GO terms and proteins, and preserves properties of both local and global structural information of the GO annotation graph. We compare our method with methods based on information content and on word embeddings, using three PPI datasets from STRING database. Experimental results demonstrate that our method is more effective than those compared methods.

BACKGROUND: Functional modules in protein-protein interaction networks (PPIN) are defined by maximal sets of functionally associated proteins and are vital to understanding cellular mechanisms and identifying disease associated proteins. Topological modules of the human proteome have been shown to be related to functional modules of PPIN. However, the effects of the weights of interactions between protein pairs and the integration of physical (direct) interactions with functional (indirect expression-based) interactions have not been investigated in the detection of functional modules of the human proteome. RESULTS: We investigated functional homogeneity and specificity of topological modules of the human proteome and validated them with known biological and disease pathways. Specifically, we determined the effects on functional homogeneity and heterogeneity of topological modules (i) with both physical and functional protein-protein interactions; and (ii) with incorporation of functional similarities between proteins as weights of interactions. With functional enrichment analyses and a novel measure for functional specificity, we evaluated functional relevance and specificity of topological modules of the human proteome. CONCLUSIONS: The topological modules ranked using specificity scores show high enrichment with gene sets of known functions. Physical interactions in PPIN contribute to high specificity of the topological modules of the human proteome whereas functional interactions contribute to high homogeneity of the modules. Weighted networks result in more number of topological modules but did not affect their functional propensity. Modules of human proteome are more homogeneous for molecular functions than biological processes.

This study aims at predicting affective states during programming using keyboard and mouse data. The article proposes and evaluates a novel set of features under programming context to predict affective states. Fourteen undergraduate participants performed three programming tasks of varying difficulties. At the completion of each task, participants reported their affective states by viewing webcam videos and screen recordings. Features extracted from keyboard and mouse logs were used to train multiple classifiers. Among trained classifiers, feedforward neural network recognized positive, neutral and negative states with 52.9% accuracy. The overall Cohen's Kappa reached 0.27. Without neutral states, the classifiers were able to differentiate positive and negative states with 74.1% accuracy and 0.48 Kappa. Our approach demonstrates improved ability of predicting self-labelled affective states of programmers from keyboard and mouse data, without using specialized sensors, and potential of emotional feedback to programmers during learning to deliver better experience.

BACKGROUND: Differential co-expression (DCX) signifies change in degree of co-expression of a set of genes among different biological conditions. It has been used to identify differential co-expression networks or interactomes. Many algorithms have been developed for single-factor differential co-expression analysis and applied in a variety of studies. However, in many studies, the samples are characterized by multiple factors such as genetic markers, clinical variables and treatments. No algorithm or methodology is available for multi-factor analysis of differential co-expression. RESULTS: We developed a novel formulation and a computationally efficient greedy search algorithm called MultiDCoX to perform multi-factor differential co-expression analysis. Simulated data analysis demonstrates that the algorithm can effectively elicit differentially co-expressed (DCX) gene sets and quantify the influence of each factor on co-expression. MultiDCoX analysis of a breast cancer dataset identified interesting biologically meaningful differentially co-expressed (DCX) gene sets along with genetic and clinical factors that influenced the respective differential co-expression. CONCLUSIONS: MultiDCoX is a space and time efficient procedure to identify differentially co-expressed gene sets and successfully identify influence of individual factors on differential co-expression.

Temporal autocorrelation present in functional magnetic resonance image (fMRI) data poses challenges to its analysis. The existing approaches handling autocorrelation in fMRI time-series often presume a specific model of autocorrelation such as an auto-regressive model. The main limitation here is that the correlation structure of voxels is generally unknown and varies in different brain regions because of different levels of neurogenic noises and pulsatile effects. Enforcing a universal model on all brain regions leads to bias and loss of efficiency in the analysis. In this paper, we propose the mixed spectrum analysis of the voxel time-series to separate the discrete component corresponding to input stimuli and the continuous component carrying temporal autocorrelation. A mixed spectral analysis technique based on M-spectral estimator is proposed, which effectively removes autocorrelation effects from voxel time-series and identify significant peaks of the spectrum. As the proposed method does not assume any prior model for the autocorrelation effect in voxel time-series, varying correlation structure among the brain regions does not affect its performance. We have modified the standard M-spectral method for an application on a spatial set of time-series by incorporating the contextual information related to the continuous spectrum of neighborhood voxels, thus reducing considerably the computation cost. Likelihood of the activation is predicted by comparing the amplitude of discrete component at stimulus frequency of voxels across the brain by using normal distribution and modeling spatial correlations among the likelihood with a conditional random field. We also demonstrate the application of the proposed method in detecting other desired frequencies.

BACKGROUND: Time delays are important factors that are often neglected in gene regulatory network (GRN) inference models. Validating time delays from knowledge bases is a challenge since the vast majority of biological databases do not record temporal information of gene regulations. Biological knowledge and facts on gene regulations are typically extracted from bio-literature with specialized methods that depend on the regulation task. In this paper, we mine evidences for time delays related to the transcriptional regulation of yeast from the PubMed abstracts. RESULTS: Since the vast majority of abstracts lack quantitative time information, we can only collect qualitative evidences of time delays. Specifically, the speed-up or delay in transcriptional regulation rate can provide evidences for time delays (shorter or longer) in GRN. Thus, we focus on deriving events related to rate changes in transcriptional regulation. A corpus of yeast regulation related abstracts was manually labeled with such events. In order to capture these events automatically, we create an ontology of sub-processes that are likely to result in transcription rate changes by combining textual patterns and biological knowledge. We also propose effective feature extraction methods based on the created ontology to identify the direct evidences with specific details of these events. Our ontologies outperform existing state-of-the-art gene regulation ontologies in the automatic rule learning method applied to our corpus. The proposed deterministic ontology rule-based method can achieve comparable performance to the automatic rule learning method based on decision trees. This demonstrates the effectiveness of our ontology in identifying rate-changing events. We also tested the effectiveness of the proposed feature mining methods on detecting direct evidence of events. Experimental results show that the machine learning method on these features achieves an F1-score of 71.43%. CONCLUSIONS: The manually labeled corpus of events relating to rate changes in transcriptional regulation for yeast is available in https://sites.google.com/site/wentingntu/data. The created ontologies summarized both biological causes of rate changes in transcriptional regulation and corresponding positive and negative textual patterns from the corpus. They are demonstrated to be effective in identifying rate-changing events, which shows the benefits of combining textual patterns and biological knowledge on extracting complex biological events.

Filter methods are often used for selection of genes in multiclass sample classification by using microarray data. Such techniques usually tend to bias toward a few classes that are easily distinguishable from other classes due to imbalances of strong features and sample sizes of different classes. It could therefore lead to selection of redundant genes while missing the relevant genes, leading to poor classification of tissue samples. In this manuscript, we propose to decompose multiclass ranking statistics into class-specific statistics and then use Pareto-front analysis for selection of genes. This alleviates the bias induced by class intrinsic characteristics of dominating classes. The use of Pareto-front analysis is demonstrated on two filter criteria commonly used for gene selection: F-score and KW-score. A significant improvement in classification performance and reduction in redundancy among top-ranked genes were achieved in experiments with both synthetic and real-benchmark data sets.

MOTIVATION: Motifs in DNA sequences often appear in degenerate form, so there has been an increased interest in computational algorithms for weak motif discovery. Probabilistic algorithms are unable to detect weak motifs while exact methods have been able to detect only short weak motifs. This article proposes an exact tree-based motif detection (TreeMotif) algorithm capable of discovering longer and weaker motifs than by the existing methods. RESULTS: TreeMotif converts the graphical representation of motifs into a tree-structured representation in which a tree that branches with nodes from every sequence represents motif instances. The method of tree construction is novel to motif discovery based on graphical representation. TreeMotif is more efficient and scalable in handling longer and weaker motifs than the existing algorithms in terms of accuracy and execution time. The performances of TreeMotif were demonstrated on synthetic data as well as on real biological data. AVAILABILITY: https://sites.google.com/site/shqssw/treemotif CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Biological networks are constantly subjected to random perturbations, and efficient feedback and compensatory mechanisms exist to maintain their stability. There is an increased interest in building gene regulatory networks (GRNs) from temporal gene expression data because of their numerous applications in life sciences. However, because of the limited number of time points at which gene expressions can be gathered in practice, computational techniques of building GRN often lead to inaccuracies and instabilities. This paper investigates the stability of sparse auto-regressive models of building GRN from gene expression data. RESULTS: Criteria for evaluating the stability of estimating GRN structure are proposed. Thereby, stability of multivariate vector autoregressive (MVAR) methods - ridge, lasso, and elastic-net - of building GRN were studied by simulating temporal gene expression datasets on scale-free topologies as well as on real data gathered over Hela cell-cycle. Effects of the number of time points on the stability of constructing GRN are investigated. When the number of time points are relatively low compared to the size of network, both accuracy and stability are adversely affected. At least, the number of time points equal to the number of genes in the network are needed to achieve decent accuracy and stability of the networks. Our results on synthetic data indicate that the stability of lasso and elastic-net MVAR methods are comparable, and their accuracies are much higher than the ridge MVAR. As the size of the network grows, the number of time points required to achieve acceptable accuracy and stability are much less relative to the number of genes in the network. The effects of false negatives are easier to improve by increasing the number time points than those due to false positives. Application to HeLa cell-cycle gene expression dataset shows that biologically stable GRN can be obtained by introducing perturbations to the data. CONCLUSIONS: Accuracy and stability of building GRN are crucial for investigation of gene regulations. Sparse MVAR techniques such as lasso and elastic-net provide accurate and stable methods for building even GRN of small size. The effect of false negatives is corrected much easier with the increased number of time points than those due to false positives. With real data, we demonstrate how stable networks can be derived by introducing random perturbation to data.

BACKGROUND: Essential events of cell development and homeostasis are revealed by the associated changes of cell morphology and therefore have been widely used as a key indicator of physiological states and molecular pathways affecting various cellular functions via cytoskeleton. Cell motility is a complex phenomenon primarily driven by the actin network, which plays an important role in shaping the morphology of the cells. Most of the morphology based features are approximated from cell periphery but its dynamics have received none to scant attention. We aim to bridge the gap between membrane dynamics and cell states from the perspective of whole cell movement by identifying cell edge patterns and its correlation with cell dynamics. RESULTS: We present a systematic study to extract, classify, and compare cell dynamics in terms of cell motility and edge activity. Cell motility features extracted by fitting a persistent random walk were used to identify the initial set of cell subpopulations. We propose algorithms to extract edge features along the entire cell periphery such as protrusion and retraction velocity. These constitute a unique set of multivariate time-lapse edge features that are then used to profile subclasses of cell dynamics by unsupervised clustering. CONCLUSIONS: By comparing membrane dynamic patterns exhibited by each subclass of cells, correlated trends of edge and cell movements were identified. Our findings are consistent with published literature and we also identified that motility patterns are influenced by edge features from initial time points compared to later sampling intervals.

Although numerous computational techniques have been applied to predict protein secondary structure (PSS), only limited studies have dealt with discovery of logic rules underlying the prediction itself. Such rules offer interesting links between the prediction model and the underlying biology. In addition, they enhance interpretability of PSS prediction by providing a degree of transparency to the predicting model usually regarded as a black box. In this paper, we explore the generation and use of C4.5 decision trees to extract relevant rules from PSS predictions modeled with two-stage support vector machines (TS-SVM). The proposed rules were derived on the RS126 data set of 126 nonhomologous globular proteins and on the PSIPRED data set of 1,923 protein sequences. Our approach has produced sets of comprehensible, and often interpretable, rules underlying the PSS predictions. Moreover, many of the rules seem to be strongly supported by biological evidence. Further, our approach resulted in good prediction accuracy, few and usually compact rules, and rules that are generally of higher confidence levels than those generated by other rule extraction techniques.

Correlation is a very widely used filter criterion for gene selection in cancer classification. However, it uses all the training samples in ranking, which may not be equally important for the classification. Using support vectors, we demonstrate that classical correlation coefficient based gene selection is biased because of the sample points away from classification margin. To remove such bias, we use only the support vectors for computation of correlation coefficient and propose a backward elimination based SVcc-RFE algorithm. The proposed method is tested on several benchmark cancer gene expression datasets and the results show improvement in classification performance compared to other state-of-the-art methods.

Independent component analysis with reference (ICA-R), is a technique to incorporate prior information about the desired sources as reference signals into the contrast function of ICA so as to form an augmented Lagrangian function under the framework of constrained ICA (cICA). The ICA-R algorithm is constructed by solving the optimization problem via Newton-like learning style. Unfortunately, this algorithm does not find a global optimum once it reaches a local optimum resulting in misconvergence that hinders the capability of ICA-R. To overcome the optimization problems with the previous methods, this paper uses an evolutionary approach to ICA-R that brings the search out of local minima and finds a global optimal solution. Experiments with synthetic signals demonstrate the validity of the proposed method.

The Complex Independent Component Analysis (CICA) which extends Independent Component Analysis (ICA) to complex signals has found applications in various fields. The ICA with Reference (ICA-R) has recently gained popularity in semi-blind separation of signals when a priori information of the desired sources are available in the form of reference signals. This paper extends the framework of ICA-R to complex signals and demonstrates the use of Complex ICA-R (CICA-R) with applications to both synthetic data and real speech data. Our experiments indicate that CICA-R is more effective than ICA, ICA-R, or CICA, in separation of complex signals when reference signals relating to source signals are available.

We introduce a probabilistic framework for building higher-order gene regulatory networks, which automatically finds the delays of regulatory interactions. A variable-order Markov chain Monte Carlo method with a new acceptance mechanism is proposed to find the optimal order and the structure of a dynamic Bayesian network (DBN). Experiments on cell cycle expression data indicate that the variable-order DBN (VDBN) better fits the data and gives biologically more plausible regulatory networks.

BACKGROUND: Weak motif discovery in DNA sequences is an important but unresolved problem in computational biology. Previous algorithms that aimed to solve the problem usually require a large amount of memory or execution time. In this paper, we proposed a fast and memory efficient algorithm, RecMotif, which guarantees to discover all motifs with specific (l, d) settings (where l is the motif length and d is the maximum number of mutations between a motif instance and the true motif). RESULTS: Comparisons with several recently proposed algorithms have shown that RecMotif is more scalable for handling longer and weaker motifs. For instance, it can solve the open challenge cases such as (40, 14) within 5 hours while the other algorithms compared failed due to either longer execution times or shortage of memory space. For real biological sequences, such as E.coli CRP, RecMotif is able to accurately discover the motif instances with (l, d) as (18, 6) in less than 1 second, which is faster than the other algorithms compared. CONCLUSIONS: RecMotif is a novel algorithm that requires only a space complexity of O(m²n) (where m is the number of sequences in the data and n is the length of the sequences).

BACKGROUND: High content screening techniques are increasingly used to understand the regulation and progression of cell motility. The demand of new platforms, coupled with availability of terabytes of data has challenged the traditional technique of identifying cell populations by manual methods and resulted in development of high-dimensional analytical methods. RESULTS: In this paper, we present sub-populations analysis of cells at the tissue level by using dynamic features of the cells. We used active contour without edges for segmentation of cells, which preserves the cell morphology, and autoregressive modeling to model cell trajectories. The sub-populations were obtained by clustering static, dynamic and a combination of both features. We were able to identify three unique sub-populations in combined clustering. CONCLUSION: We report a novel method to identify sub-populations using kinetic features and demonstrate that these features improve sub-population analysis at the tissue level. These advances will facilitate the application of high content screening data analysis to new and complex biological problems.

A novel approach for gene classification, which adopts codon usage bias as input feature vector for classification by support vector machines (SVM) is proposed. The DNA sequence is first converted to a 59-dimensional feature vector where each element corresponds to the relative synonymous usage frequency of a codon. As the input to the classifier is independent of sequence length and variance, our approach is useful when the sequences to be classified are of different lengths, a condition that homology-based methods tend to fail. The method is demonstrated by using 1,841 Human Leukocyte Antigen (HLA) sequences which are classified into two major classes: HLA-I and HLA-II; each major class is further subdivided into sub-groups of HLA-I and HLA-II molecules. Using codon usage frequencies, binary SVM achieved accuracy rate of 99.3% for HLA major class classification and multi-class SVM achieved accuracy rates of 99.73% and 98.38% for sub-class classification of HLA-I and HLA-II molecules, respectively. The results show that gene classification based on codon usage bias is consistent with the molecular structures and biological functions of HLA molecules.

Information on secondary structures of amino acid residues in proteins provides valuable clues for the prediction of their 3-D structure and function. Although numerous computational techniques have been applied to predict protein secondary structure (PSS), only limited studies have dealt with discovery of logic rules underlying the prediction itself. Such rules offer interesting links between the prediction model and the underlying biology. In addition, they enhance interpretability of PSS prediction by providing a degree of transparency to the predicting model usually regarded as a black-box. In this paper, we explore the generation and use of C 4.5 decision trees to extract relevant rules from PSS predictions modeled with two-stage support vector machines (TS-SVM). Our approach has produced sizable sets of comprehensible, and often interpretable, rules underlying the PSS predictions. Moreover, many of the rules seem to be strongly supported by biological evidence. Further, our approach resulted in good prediction accuracy, few and usually compact rules, and rules that are generally of higher confidence levels than those generated by other rule extraction techniques. The proposed rules were derived and tested on the RS126 dataset of 126 nonhomologous globular proteins.

There has been an increasing interest recently in identifying subcellular proteins from cellular images in order to understachind subcellular activities of cells. However, accuracies of the prediction tend to decrease with the number of protein subcellular localization classes. Therefore in this paper, we introduce a multiple-cell model with a higher-order Markov random fields (MRF) to combine predictions on multiple cells to make inferences on protein localizations of individual cells. The proposed method showed a significant improvement in discrimination of protein subcellular localization patterns over the predictions by single cells. We also introduce structure learning of MRF, which indeed enhanced the predictions especially when the number of cells in the model becomes large.

Extracting brain rhythms from EEG signals has many applications including Brain Computer Interfacing. Here, we demonstrate how ICA with Reference (ICA-R) is used to extract brain rhythms, using appropriate reference signals. In particular, we evaluate four criteria for generating reference signals to use with ICA-R. We demonstrate the performance of these techniques in extracting mu and beta rhythms from two real EEG datasets. The results indicate that ICA-R can be effectively used for extracting brain rhythms. The blind source separation technique decomposing autocorrelation for extracting reference signals outperformed other methods.

This paper unifies our earlier work on detection of brain activation (Rajapakse and Piyaratna, 2001) and connectivity (Rajapakse and Zhou, 2007) in a probabilistic framework for analyzing effective connectivity among activated brain regions from functional magnetic resonance imaging (fMRI) data. Interactions among brain regions are expressed by a dynamic Bayesian network (DBN) while contextual dependencies within functional images are formulated by a Markov random field. The approach simultaneously considers both the detection of brain activation and the estimation of effective connectivity and does not require a priori model of connectivity. Experimental results show that the present approach outperforms earlier fMRI analysis techniques on synthetic functional images and robustly derives brain connectivity from real fMRI data.

Pseudogenes are formed by either gene duplication or retro transposition and yet unknown to express any RNA or produce any protein, which may be due to some defects in their structure. Because of the non-functional nature, pseudogenes are considered important resources in the study of evolutionary history and phylogenetic comparison of genomes. Psuedogenes whose structure is similar to the normal genes, pose problems in gene annotation and interfere with PCR or hybridization experiments. In this paper, we compare structural and functional properties of pseudogenes of human and mouse. It was found that 3,277 pseudogenes of humans which had conserved regions in pseudogenes of mouse shared the same number of chromosomes. Human and mouse pseudogenes are very similar to each other based on the effective codon usage and fraction of codons having guanine or cytosine at the third codon position. However, the proportion of GC content and lengths of base pairs were different. The parent genes or proteins which have more number of pseudogenes may be considered to be evolving more quickly showing more variability. Further ribosomal proteins, binding proteins and receptors had more number of pseudogenes than the other proteins

The ten papers in this special section focus on computational intelligence approaches on computational biology and bioinformatics.

Knowledge of protein-protein interaction sites is vital to determine proteins' function and involvement in different pathways. Though a wide variety of methods has been proposed over the recent years in order to predict protein-protein interface residues, mainly based on single amino acid sequence inputs, each has its own drawbacks and limitations. We propose to use support vector machines (SVM) to predict protein-protein interface residues by using the information derived from the position-specific scoring matrices (PSSMs) generated from PSI-BLAST profiles and accessible surface areas. The present approach achieved overall prediction accuracy of 73.8% for 77 individulal proteins collected from the Protein Data Bank, which is better than the previously reported accuracies

The codon usage patterns of 2,552 major histocompatibility complex (MHC) sequences from 33 primate species, and the consequent subsets of sequences obtained by removing species with most abundant sequences was observed. The correlation between function and species with regards to MHC codon usage patterns was analyzed using cluster analysis and support vector machines (SVMs). The results show that gene function is the major factor, while species is the minor factor correlated to codon usage bias, but their interactions complicate the codon usage pattern. When the weight of the factor of species increases, the accuracy rate of classification dropped accordingly. The factors of gene function and species can be adopted as feature vectors in the field of gene classification and phylogenetic studies respectively. As the input of codon usage to the classifier is independent of sequence length and variance, our approach is useful when the sequences to be analyzed are of different lengths, a condition where classic homology-based approaches tend to be difficult. To focus on the phylogenetic features of the MHC sequences through codon usage analysis, we must try to minimize or even eliminate the influence of gene function

This paper presents a conditional random field (CRF) approach to fuse contextual dependencies in functional magnetic resonance imaging (fMRI) data for the detection of brain activation. The interactions among both activation (activated/inactive) labels and observed data of brain voxels are unified in a probabilistic framework based on the CRF, where the interaction strength can be adaptively adjusted in terms of the data similarity of neighboring sites. Compared to earlier detection methods, including statistical parametric mapping and Markov random field, the proposed method avoids the suppression of high frequency information and relaxes the strong assumption of conditional independence of observed data. Experimental results show that the proposed approach effectively integrates contextual constraints within the detection process and robustly detects brain activities from fMRI data.

This paper delves into the application of a new technique, independent component analysis with reference (ICA-R), in order to obtain the components of ERPs from the EEC data. The technique is tested on available EEC data on a picture matching task performed by alcoholic and nonalcoholic subjects. The accuracy, efficiency, and usefulness of ICA-R are demonstrated in extracting the relevant components of single-trial ERPs.

Several candidate genes affect Parkinson's disease in a variety of ways. A comparative analysis is performed using orthologues of these genes in other vertebrate species, such as chimp, mouse, rat, chicken, fugu, and tetraodon. The analysis reveals the presence of transmembrane regions and signal peptides in several sequences of some species, which provides a better understanding of the variability of structural and functional aspects of these genes in different species.

We present a technique to encode the inputs to neural networks for the detection of signals in genomic sequences. The encoding is based on lower-order Markov models which incorporate known biological characteristics in genomic sequences. The neural networks then learn intrinsic higher-order dependencies of nucleotides at the signal sites. We demonstrate the efficacy of the Markov encoding method in the detection of three genomic signals, namely, splice sites, transcription start sites, and translation initiation sites.

This paper presents the technique of constrained independent component analysis (cICA) and demonstrates two applications, less-complete ICA, and ICA with reference (ICA-R). The cICA is proposed as a general framework to incorporate additional requirements and prior information in the form of constraints into the ICA contrast function. The adaptive solutions using the Newton-like learning are proposed to solve the constrained optimization problem. The applications illustrate the versatility of the cICA by separating subspaces of independent components according to density types and extracting a set of desired sources when rough templates are available. The experiments using face images and functional MR images demonstrate the usage and efficacy of the cICA.

Bioinformatics techniques to relative solvent accessibility (RSA) prediction are mostly single-stage approaches; they predict solvent accessibility of proteins by taking into account only the information available in amino acid sequences. We propose to use support vector machines (SVMs) as a second stage following the existing single-stage approaches for RSA prediction problem to improve the accuracy. The purpose of the second stage is to capture the contextual relationship of solvent accessibility elements in a neighborhood in determining the solvent accessibility at a particular site. We demonstrate our approach by introducing SVMs to the output of single-stage SVM classifier. Two-stage SVM approach achieves accuracies up to 90.4% and 90.2% on the Manesh dataset of 215 protein structures and the RS126 dataset of 126 nonhomologous globular proteins, respectively, which are better than the highest reported scores on both datasets to date.

Feature selection is a commonly addressed problem in classification. In gene expression-based cancer classification, a large number of genes in conjunction with a small number of samples makes the gene selection problem more important but also more challenging. Support vector machine as a popular classification algorithm, has been successfully used in SVM-RFE method for gene selection. This paper proposes a variant of SVM-RFE to do gene selection for cancer classification with expression data. Multiple support vector machine classifiers from a leave-one-out procedure are used to compute the feature ranking scores. The numerical experiments also show the good and stable performance of the proposed method.

Low-order Markov models are insufficient to represent hidden and complex features surrounding translation initiation sites (TISs). We present a neural network approach for detecting TISs of eukaryotes that combines lower-order models carrying biological knowledge, with nonlinearity, to capture higher-order nucleotide correlations at TISs and in the surrounding coding and noncoding regions. The model consists of Markov chain models of low-order, a protein encoding model, neural networks, and a ribosome scanning model. The Markov models capture the local interaction of coding and noncoding regions while the protein encoding model utilizes evolutionary information of proteins to encode the downstream coding sequence of the TIS. With the incorporation of ribosome scanning model, the present method allows for incorporating the biological contextual information of potential TIS into the neural network based prediction system, thereby increasing number of correct recognitions considerably. A 3-fold cross-validation evaluation on the Pedersen and Nielsen dataset yielded an average 93.8% of sensitivity and 96.9% of specificity, indicating superior accuracy of the present system to the previous highest measures of 88.5% of sensitivity and 963% of specificity.

Comparative analysis of evolutionary rates of different species provides us with informative results as to how nucleotides mutate in species with different divergence times. Relative rate test is frequently deployed to test rates of synonymous and nonsynonymous substitutions. We are interested to use the test on a set of genes which are related to the Parkinson's Disease. Comparative genomics of human, chimp, mouse, chicken and fugu revealed most of the genes did not provide evidence of a rate heterogeneity, thereby inferring most of them are undergoing neutral or nearly neutral mutation. This allows us to get a better understanding of the evolutionary traits of genes implicated in Parkinson's Disease.

Several algorithms have been developed for motif recognition in the past few years, superior in some sense over others, yet not a single one was declared to be the "best". Some of the well recognized algorithms are based on heuristic methods, such as Gibbs sampling and expectation maximization, and enumeration methods, such as Oligo-analysis. However, the inability to solve the "Challenge Problem" in motif recognition showed the drawbacks of the existing heuristic and enumeration methods. Two new algorithms were developed to resolve this problem but still, they suffered from time and space expense and the problem of local optima. We proposed a new algorithm which can solve the challenge problem with better performance even in very long sequences by applying dynamic programming for path searching in a graph and scanning with the consensus sequence to eliminate faked motif instances.

We summarize some of the key research areas in brain sciences and neural networks that have recently been or are being worked on by researchers in Singapore. Researchers in Singapore are developing theory of neural networks, notably improved radial basis function networks, fuzzy neural networks, and fast learning neural networks. Applications of neural networks include bioinformatics, multimedia, data mining, and communications. Researchers are also working with neurophysiologists on functional brain imaging and brain disease analysis.

This paper proposes a novel approach for analysis of brain connectivity shown in functional MRI (fMRI), using graphical models. Structural equation modeling (SEM) is currently used to model neural systems by using partial covariance values, which is only able to affirm or refute functional connectivity of a previously known anatomical model or select the best fit model from a set of a priori models. Our approach is exploratory in the sense that it does not require a priori model such as an anatomical model. The SEM uses covariances which describe only second order behavior of a network while conditional probabilities used in graphical models, in theory, describe the complete behavior of a network. It renders the interactions among brain regions with conditional densities and allows simulation of disconnectivity of neural systems.

Extracting the human brain from magnetic resonance head scans is difficult because of its highly convoluted and nonuniform geometry. A technique based on Non-Uniform Rational B-Splines (NURBS) surfaces and energy minimizing deformable models to extract and visualize the brain surface patterns accurately from magnetic resonance head scans is presented. The weighting parameter that comes with the NURBS definition is explored to attract the surface into regions showing high curvature. The weight at each control point is adjusted automatically according to the curvature properties of the evolving surface. This process facilitates a deformable model with increased local flexibility that adapts to complex geometrical features of the brain surface. The results show that the proposed model is capable of capturing the correct brain surface with a higher accuracy than the existing techniques.

Frequency domain analysis of event-related and block-design functional MR imaging (fMRI) time-series is presented, assuming a linear model of fMR response. The frequency domain least-square estimation is performed in order to evaluate hemodynamic parameters of the brain. The usefulness and limitations of the frequency domain analysis of fMRI time-series are highlighted.

A probabilistic approach combining a hidden Markov model and neural networks is implemented to identify different functional entities in nucleotide sequences. This approach augments the Hidden Markov model probability parameters by using the outputs of neural networks. It is designed to capture the compositional properties of complex genes and thereby achieves low error rates and high correlation coefficient measures. Initial experiments demonstrate that the predictive efficiency of the model is considerably higher than the existing models of gene finding.

A fuzzy neural approach is presented to segment subcortical structures, such as amygdala, caudate, putamen, hippocampus and thalamus, automatically, from MR head scans. Both the position and intensity features of these structures are used by fuzzy neural network and clustering techniques, incorporating a priori knowledge from neuroanatomy. Information of tissue classes and the positions of voxels are fused together to make decisions on them belonging to specific anatomical structures of the brain. Experimental results are presented to demonstrate the accuracy and use of the technique.

We propose a novel approach to predict protein secondary structure by combining different types of GOR (Garnier, Osguthorpe, and Robson) classifiers with Support Vector Machines (SVMs). The new prediction scheme achieves an accuracy of 69.3% when using the sevenfold cross validation on a database of 126 nonhomologous globular proteins. Applying the method to multiple sequence alignments of homologous proteins significantly increases the prediction accuracy to 72.1%. We show that it is possible to obtain a higher accuracy with combined classifiers than GOR classifiers or Support Vector Machines alone, in protein secondary structure prediction.

An image segmentation scheme is presented which combines the desired features of region growing and watershed transform into a single framework. The segmentation process is guided by the edgeness profile of the image analogous to a topographic relief in which local minima act as seeds for segments. Growing regions are allowed to split in the catchment basins of the edgeness profile and merge at watersheds depending on the similarities of their regional intensity variations. The experiments show that the proposed segmentation scheme produces homogeneous regions with accurate boundaries and better segmentations than those produced by region growing or watershed transform followed by region merging.

Application of "blind" processing, based on unsupervised learning in principal, for the analysis of functional brain images does not assume any a priori information on the signals generated in the brain. We describe blind processing strategies with an emphasis on Independent Component Analysis (ICA) for the analysis of functional Magnetic Resonance (fMR) images. We present a semi-blind processing technique, referred to as ICA with Reference, to overcome some limitations of pure blind-processing of fMRI data. Application of ICA and ICA-R on real fMRI datasets are presented.

A statistical approach based on non-uniform rational b-splines (NURBS) to detect and visualize the variability associated between the brain surfaces of male and female adults is presented. A deformable NURBS surface is used initially to segment the brains from normalized magnetic resonance (MR) images. The NURBS surface of an adult brain is then deformed to converge onto a representative subject image. Significant deformation of surface points are captured by statistical analysis of the deformation vectors. To detect significant variability in sulci and gyri patterns, the Laplacian curvature of the significantly deformed points are computed. Results are demonstrated using MR head scans of 20 normal male and female adults.

The paper presents a novel technique of constrained independent component analysis (CICA) to introduce constraints into the clas(cid:173) sical ICA and solve the constrained optimization problem by using Lagrange multiplier methods. This paper shows that CICA can be used to order the resulted independent components in a specific manner and normalize the demixing matrix in the signal separation procedure. It can systematically eliminate the ICA's indeterminacy on permutation and dilation. The experiments demonstrate the use of CICA in ordering of independent components while providing normalized demixing processes. Keywords: Independent component analysis, constrained indepen(cid:173) dent component analysis, constrained optimization, Lagrange mul(cid:173) tiplier methods

A statistical model is presented that represents the distributions of major tissue classes in single-channel magnetic resonance (MR) cerebral images. Using the model, cerebral images are segmented into gray matter, white matter, and cerebrospinal fluid (CSF). The model accounts for random noise, magnetic field inhomogeneities, and biological variations of the tissues. Intensity measurements are modeled by a finite Gaussian mixture. Smoothness and piecewise contiguous nature of the tissue regions are modeled by a three-dimensional (3-D) Markov random field (MRF). A segmentation algorithm, based on the statistical model, approximately finds the maximum a posteriori (MAP) estimation of the segmentation and estimates the model parameters from the image data. The proposed scheme for segmentation is based on the iterative conditional modes (ICM) algorithm in which measurement model parameters are estimated using local information at each site, and the prior model parameters are estimated using the segmentation after each cycle of iterations. Application of the algorithm to a sample of clinical MR brain scans, comparisons of the algorithm with other statistical methods, and a validation study with a phantom are presented. The algorithm constitutes a significant step toward a complete data driven unsupervised approach to segmentation of MR images in the presence of the random noise and intensity inhomogeneities.

A neural architecture is presented for fusion of multisensory information at the feature level. The inputs to the network are in the form of binary edge patterns from multiple sensors. Each input is processed by a hierarchical neural structure similar to the forward path of MARA, which was previously proposed by the authors (1990). Since the neurons at higher levels are insensitive to distortion, noise, scaling, and displacement of the activities at lower levels, the activities of higher-level neurons in these networks processing different sensor information can be combined. The proposed architecture consists of a fusion path to carry the combined information to indicate the final decision. The fusion architecture is used to recognize the traces of trained patterns in low-quality images from different sensors.>

Presents a hierarchical neural network architecture which can achieve multisensor data fusion at feature level by combining the coherent activities of higher-level neurons in the architecture. The architecture learns very quickly and is highly stable and plastic. Initial results show promise. The authors utilize the fusion architecture to recognize traces of trained patterns in low-quality images after fusion through the architecture

The multilayer adaptive resonance architecture (MARA) is a highly stable and plastic self-organizing neural network which is capable of recognizing, reconstructing, and segmenting the traces of previously learned binary patterns. The recognition and reconstruction properties of the network are invariant with respect to distortion, noise, translation, scaling and partial rotation of the original training patterns. If multiple traces of trained patterns are simultaneously presented, the network can separate and reconstruct the traces separately. A description is given of MARA and an image segmentation scheme using MARA. MARA's ability to segment some biological structures in high-speed synchronous volume CT images of the heart is demonstrated