Multimodal Large Language Models (MLLMs) have achieved impressive performance across a variety of vision–language tasks. However, their internal working mechanisms remain largely underexplored. In this work, we introduce FG-Tracer, a framework designed to analyze the information flow between visual and textual modalities in MLLMs in free-form generation. Notably, our numerically stabilized computational method enables the first systematic analysis of multimodal information flow in underexplored domains such as image captioning and chain-of-thought (CoT) reasoning. We apply FG-Tracer to three state-of-the-art MLLMs—LLaVA 1.5, LLaMA 3.2-Vision, and Qwen 2.5-VL—across three vision–language benchmarks—TextVQA, COCO 2014, and ChartQA—and we conduct a word-level analysis of multimodal integration. Our findings uncover distinct patterns of multimodal fusion across models and tasks, demonstrating that fusion dynamics are both model- and task-dependent. Overall, FG-Tracer offers a robust methodology for probing the internal mechanisms of MLLMs in free-form settings, providing new insights into their multimodal reasoning strategies. Our source code is publicly available at https://github.com/AImageLab-zip/FG-TRACER

Cone-Beam computed tomography (CBCT) is a standard imaging modality in orofacial and dental practices, providing essential 3D volumetric imaging of anatomical structures, including jawbones, teeth, sinuses, and neurovascular canals. Accurately segmenting these structures is fundamental to numerous clinical applications, such as surgical planning and implant placement. However, manual segmentation of CBCT scans is time-intensive and requires expert input, creating a demand for automated solutions through deep learning. Effective development of such algorithms relies on access to large, well-annotated datasets, yet current datasets are often privately stored or limited in scope and considered structures, especially concerning 3D annotations. This paper proposes ToothFairy2, a comprehensive, publicly accessible CBCT dataset with voxel-level 3D annotations of 42 distinct classes corresponding to maxillofacial structures. We validate the dataset by benchmarking state-of-the-art neural network models, including convolutional, transformer-based, and hybrid Mamba-Based architectures, to evaluate segmentation performance across complex anatomical regions. Our work also explores adaptations to the nnU-Net framework to optimize multi-class segmentation for maxillofacial anatomy. The proposed dataset provides a fundamental resource for advancing maxillofacial segmentation and supports future research in automated 3D image analysis in digital dentistry.

Foundation models serve as the backbone for numerous specialized models developed through fine-tuning. However, when the underlying pretrained model is updated or retrained (e.g., on larger and more curated datasets), the fine-tuned model becomes obsolete, losing its utility and requiring retraining. This raises the question: is it possible to transfer fine-tuning to a new release of the model? In this work, we investigate how to transfer fine-tuning to a new checkpoint without having to re-train, in a data-free manner. To do so, we draw principles from model re-basin and provide a recipe based on weight permutations to re-base the modifications made to the original base model, often called task vector. In particular, our approach tailors model re-basin for Transformer models, taking into account the challenges of residual connections and multi-head attention layers. Specifically, we propose a two-level method rooted in spectral theory, initially permuting the attention heads and subsequently adjusting parameters within select pairs of heads. Through extensive experiments on visual and textual tasks, we achieve the seamless transfer of fine-tuned knowledge to new pre-trained backbones without relying on a single training step or datapoint. Code is available at https://github.com/aimagelab/TransFusion.

Hepatocellular carcinoma (HCC) is a complex and heterogeneous disease that poses significant challenges for diagnosis and treatment. Recent advancements in medical imaging and foundation models have provided new opportunities to enhance our understanding of HCC. However, these foundation models appear to lack the generalization capabilities when applied to diverse HCC cohorts and prediction tasks. This paper provides a critical review of the current progress and limitations in the generalization of foundation models in HCC pathology. It leverages numerous foundation models and multiple instance learning methods on publicly available TCGA data and a private cohort to evaluate tumor grade and overall survival estimation. To simulate real-world clinical scenarios, we assess the adaptation of foundation models under a data scarcity regime, analyzing their robustness to domain shifts. Results indicate that while foundation models exhibit stable but suboptimal performance across both tasks, their representation power for less common applications such as HCC remains limited, likely due to under-representation in the pretraining datasets. Furthermore, a substantial performance instability in zero-shot analysis highlights the necessity of fine-tuning on the target domain for reliable deployment in clinical settings. These findings underscore the challenges of applying foundation models in specialized medical domains and emphasize the need for improved adaptation strategies for histopathology applications.

Deep vision models often rely on biases learned from spurious correlations in datasets. To identify these biases, methods that interpret high-level, human-understandable concepts are more effective than those relying primarily on low-level features like heatmaps. A major challenge for these concept-based methods is the lack of image annotations indicating potentially bias-inducing concepts, since creating such annotations requires detailed labeling for each dataset and concept, which is highly labor-intensive. We present CUBIC (Concept embeddings for Unsupervised Bias IdentifiCation), a novel method that automatically discovers interpretable concepts that may bias classifier behavior. Unlike existing approaches, CUBIC does not rely on predefined bias candidates or examples of model failures tied to specific biases, as these are not always available in the data. Instead, it utilizes image-text latent space and linear classifier probes to examine how the latent representation of a superclass label—shared by all instances in the dataset—is influenced by the presence of a concept. By measuring these shifts against the normal vector to the classifier’s decision boundary, CUBIC identifies concepts that significantly influence model predictions. Our experiments demonstrate that CUBIC effectively uncovers previously unknown biases using Vision-Language Models (VLMs) without requiring the samples in the dataset where the classifier underperforms or prior knowledge of potential biases.

Continual Learning (CL) has emerged as a paramount area in Artificial Intelligence (AI) because of its ability to learn multiple tasks sequentially without significant performance degradation. Despite the growing interest in CL frameworks, a critical aspect must be addressed: the inherent biases within training data. In this work, we show that, if overlooked, these biases can significantly impair the efficacy of continual learning models by inducing reliance on suboptimal shortcuts during data stream and memory retention, exacerbating catastrophic forgetting. In response, we present Learning without Shortcuts (LwS), which sets forth two primary objectives: (i) to identify and mitigate the exploitation of spurious correlations within the data stream and (ii) to develop a novel mechanism that constructs a fair memory buffer used in replay-based CL strategies. Our buffer construction strategy exploits the model confidence in a given example to balance the portion of samples per class, hence their contribution when replay activates. Unlike existing methods, LwS is agnostic to protected attributes, and results highlight that the proposed solution is indeed resilient to spurious correlations in CL settings. Code is available at https://github.com/aimagelab/mammoth.

This paper tackles the domain of multimodal prompting for visual recognition, specifically when dealing with missing modalities through multimodal Transformers. It presents two main contributions: (i) we introduce a novel prompt learning module which is designed to produce sample-specific prompts and (ii) we show that modalityagnostic prompts can effectively adjust to diverse missing modality scenarios. Our model, termed SCP, exploits the semantic representation of available modalities to query a learnable memory bank, which allows the generation of prompts based on the semantics of the input. Notably, SCP distinguishes itself from existing methodologies for its capacity of self-adjusting to both the missing modality scenario and the semantic context of the input, without prior knowledge about the specific missing modality and the number of modalities. Through extensive experiments, we show the effectiveness of the proposed prompt learning framework and demonstrate enhanced performance and robustness across a spectrum of missing modality cases. Our source code is available at https://github.com/vittoriopipoli/SCP_WACV2025.

The failures of Deep Networks can sometimes be ascribed to biases in the data or algorithmic choices. Existing debiasing approaches exploit prior knowledge to avoid unintended solutions; we acknowledge that, in real-world settings, it could be unfeasible to gather enough prior information to characterize the bias, or it could even raise ethical considerations. We hence propose a novel debiasing approach, termed ClusterFix, which does not require any external hint about the nature of biases. Such an approach alters the standard empirical risk minimization and introduces a per-example weight, encoding how critical and far from the majority an example is. Notably, the weights consider how difficult it is for the model to infer the correct pseudo-label, which is obtained in a self-supervised manner by dividing examples into multiple clusters. Extensive experiments show that the misclassification error incurred in identifying the correct cluster allows for identifying examples prone to bias-related issues. As a result, our approach outperforms existing methods on standard benchmarks for bias removal and fairness.

The usage of Multi Instance Learning (MIL) for classifying Whole Slide Images (WSIs) has recently increased. Due to their gigapixel size, the pixel-level annotation of such data is extremely expensive and time-consuming, practically unfeasible. For this reason, multiple automatic approaches have been raised in the last years to support clinical practice and diagnosis. Unfortunately, most state-of-the-art proposals apply attention mechanisms without considering the spatial instance correlation and usually work on a single-scale resolution. To leverage the full potential of pyramidal structured WSI, we propose a graph-based multi-scale MIL approach, DAS-MIL. Our model comprises three modules: i) a self-supervised feature extractor, ii) a graph-based architecture that precedes the MIL mechanism and aims at creating a more contextualized representation of the WSI structure by considering the mutual (spatial) instance correlation both inter and intra-scale. Finally, iii) a (self) distillation loss between resolutions is introduced to compensate for their informative gap and significantly improve the final prediction. The effectiveness of the proposed framework is demonstrated on two well-known datasets, where we outperform SOTA on WSI classification, gaining a +2.7% AUC and +3.7% accuracy on the popular Camelyon16 benchmark.

We introduce Neuro-Symbolic Continual Learning, where a model has to solve a sequence of neuro-symbolic tasks, that is, it has to map sub-symbolic inputs to high-level concepts and compute predictions by reasoning consistently with prior knowledge. Our key observation is that neuro-symbolic tasks, although different, often share concepts whose semantics remains stable over time. Traditional approaches fall short: existing continual strategies ignore knowledge altogether, while stock neuro-symbolic architectures suffer from catastrophic forgetting. We show that leveraging prior knowledge by combining neuro-symbolic architectures with continual strategies does help avoid catastrophic forgetting, but also that doing so can yield models affected by reasoning shortcuts. These undermine the semantics of the acquired concepts, even when detailed prior knowledge is provided upfront and inference is exact, and in turn continual performance. To overcome these issues, we introduce COOL, a COncept-level cOntinual Learning strategy tailored for neuro-symbolic continual problems that acquires high-quality concepts and remembers them over time. Our experiments on three novel benchmarks highlights how COOL attains sustained high performance on neuro-symbolic continual learning tasks in which other strategies fail.

Messenger RNA (mRNA) has an essential role in the protein production process. Predicting mRNA expression levels accurately is crucial for understanding gene regulation, and various models (statistical and neural network-based) have been developed for this purpose. A few models predict mRNA expression levels from the DNA sequence, exploiting the DNA sequence and gene features (e.g., number of exons/introns, gene length). Other models include information about long-range interaction molecules (i.e., enhancers/silencers) and transcriptional regulators as predictive features, such as transcription factors (TFs) and small RNAs (e.g., microRNAs - miRNAs). Recently, a convolutional neural network (CNN) model, called Xpresso, has been proposed for mRNA expression level prediction leveraging the promoter sequence and mRNAs' half-life features (gene features). To push forward the mRNA level prediction, we present miREx, a CNN-based tool that includes information about miRNA targets and expression levels in the model. Indeed, each miRNA can target specific genes, and the model exploits this information to guide the learning process. In detail, not all miRNAs are included, only a selected subset with the highest impact on the model. MiREx has been evaluated on four cancer primary sites from the genomics data commons (GDC) database: lung, kidney, breast, and corpus uteri. Results show that mRNA level prediction benefits from selected miRNA targets and expression information. Future model developments could include other transcriptional regulators or be trained with proteomics data to infer protein levels.

Recent developments have resulted in multiple techniques trying to explain how deep neural networks achieve their predictions. The explainability maps provided by such techniques are useful to understand what the network has learned and increase user confidence in critical applications such as the medical field or autonomous driving. Nonetheless, they typically have very low resolutions, severely limiting their capability of identifying finer details or multiple subjects. In this paper we employ an encoder-decoder architecture with skip connection known as U-Net, originally developed for segmenting medical images, as an image classifier and we show that state of the art explainable techniques applied to U-Net can generate pixel level explanation maps for images of any resolution.

MOTIVATION: Computer-aided analysis of biological images typically requires extensive training on large-scale annotated datasets, which is not viable in many situations. In this paper, we present Generative Adversarial Network Discriminator Learner (GAN-DL), a novel self-supervised learning paradigm based on the StyleGAN2 architecture, which we employ for self-supervised image representation learning in the case of fluorescent biological images. RESULTS: We show that Wasserstein Generative Adversarial Networks enable high-throughput compound screening based on raw images. We demonstrate this by classifying active and inactive compounds tested for the inhibition of SARS-CoV-2 infection in two different cell models: the primary human renal cortical epithelial cells (HRCE) and the African green monkey kidney epithelial cells (VERO). In contrast to previous methods, our deep learning-based approach does not require any annotation, and can also be used to solve subtle tasks it was not specifically trained on, in a self-supervised manner. For example, it can effectively derive a dose-response curve for the tested treatments. AVAILABILITY AND IMPLEMENTATION: Our code and embeddings are available at https://gitlab.com/AlesioRFM/gan-dl StyleGAN2 is available at https://github.com/NVlabs/stylegan2 .

BACKGROUND: The function of non-coding RNA sequences is largely determined by their spatial conformation, namely the secondary structure of the molecule, formed by Watson-Crick interactions between nucleotides. Hence, modern RNA alignment algorithms routinely take structural information into account. In order to discover yet unknown RNA families and infer their possible functions, the structural alignment of RNAs is an essential task. This task demands a lot of computational resources, especially for aligning many long sequences, and it therefore requires efficient algorithms that utilize modern hardware when available. A subset of the secondary structures contains overlapping interactions (called pseudoknots), which add additional complexity to the problem and are often ignored in available software. RESULTS: We present the SeqAn-based software LaRA 2 that is significantly faster than comparable software for accurate pairwise and multiple alignments of structured RNA sequences. In contrast to other programs our approach can handle arbitrary pseudoknots. As an improved re-implementation of the LaRA tool for structural alignments, LaRA 2 uses multi-threading and vectorization for parallel execution and a new heuristic for computing a lower boundary of the solution. Our algorithmic improvements yield a program that is up to 130 times faster than the previous version. CONCLUSIONS: With LaRA 2 we provide a tool to analyse large sets of RNA secondary structures in relatively short time, based on structural alignment. The produced alignments can be used to derive structural motifs for the search in genomic databases.

Due to the current high availability of omics, data-driven biology has greatly expanded, and several papers have reviewed state-of-the-art technologies. Nowadays, two main types of investigation are available for a multi-omics dataset: extraction of relevant features for a meaningful biological interpretation and clustering of the samples. In the latter case, a few reviews refer to some outdated or no longer available methods, whereas others lack the description of relevant clustering metrics to compare the main approaches. This work provides a general overview of the major techniques in this area, divided into four groups: graph, dimensionality reduction, statistical and neural-based. Besides, eight tools have been tested both on a synthetic and a real biological dataset. An extensive performance comparison has been provided using four clustering evaluation scores: Peak Signal-to-Noise Ratio (PSNR), Davies-Bouldin(DB) index, Silhouette value and the harmonic mean of cluster purity and efficiency. The best results were obtained by using the dimensionality reduction, either explicitly or implicitly, as in the neural architecture.

It is estimated that oncogenic gene fusions cause about 20% of human cancer morbidity. Identifying potentially oncogenic gene fusions may improve affected patients' diagnosis and treatment. Previous approaches to this issue included exploiting specific gene-related information, such as gene function and regulation. Here we propose a model that profits from the previous findings and includes the microRNAs in the oncogenic assessment. We present ChimerDriver, a tool to classify gene fusions as oncogenic or not oncogenic. ChimerDriver is based on a specifically designed neural network and trained on genetic and post-transcriptional information to obtain a reliable classification. The designed neural network integrates information related to transcription factors, gene ontologies, microRNAs and other detailed information related to the functions of the genes involved in the fusion and the gene fusion structure. As a result, the performances on the test set reached 0.83 f1-score and 96% recall. The comparison with state-of-the-art tools returned comparable or higher results. Moreover, ChimerDriver performed well in a real-world case where 21 out of 24 validated gene fusion samples were detected by the gene fusion detection tool Starfusion. ChimerDriver integrates transcriptional and post-transcriptional information in an ad-hoc designed neural network to effectively discriminate oncogenic gene fusions from passenger ones. ChimerDriver source code is freely available at https://github.com/martalovino/ChimerDriver.

Incremental learning involves Machine Learning paradigms that dynamically adjust their previous knowledge whenever new training samples emerge. To address the problem of multi-task incremental learning without storing any samples of the previous tasks, the so-called Expert Gate paradigm was proposed, which consists of a Gate and a downstream network of task-specific CNNs, a.k.a. the Experts. The gate forwards the input to a certain expert, based on the decision made by a set of autoencoders. Unfortunately, as a CNN is intrinsically incapable of dealing with inputs of a class it was not specifically trained on, the activation of the wrong expert will invariably end into a classification error. To address this issue, we propose a probabilistic extension of the classic Expert Gate paradigm. Exploiting the prediction uncertainty estimations provided by Bayesian Convolutional Neural Networks (B-CNNs), the proposed paradigm is able to either reduce, or correct at a later stage, wrong decisions of the gate. The goodness of our approach is shown by experimental comparisons with state-of-the-art incremental learning methods.

MOTIVATION: Fusion genes are both useful cancer biomarkers and important drug targets. Finding relevant fusion genes is challenging due to genomic instability resulting in a high number of passenger events. To reveal and prioritize relevant gene fusion events we have developed FUsionN Gene Identification toolset (FUNGI) that uses an ensemble of fusion detection algorithms with prioritization and visualization modules. RESULTS: We applied FUNGI to an ovarian cancer dataset of 107 tumor samples from 36 patients. Ten out of 11 detected and prioritized fusion genes were validated. Many of detected fusion genes affect the PI3K-AKT pathway with potential role in treatment resistance. AVAILABILITYAND IMPLEMENTATION: FUNGI and its documentation are available at https://bitbucket.org/alejandra_cervera/fungi as standalone or from Anduril at https://www.anduril.org. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Tumors are composed by a number of cancer cell subpopulations (subclones), characterized by a distinguishable set of mutations. This phenomenon, known as intra-tumor heterogeneity (ITH), may be studied using Copy Number Aberrations (CNAs). Nowadays ITH can be assessed at the highest possible resolution using single-cell DNA (scDNA) sequencing technology. Additionally, single-cell CNA (scCNA) profiles from multiple samples of the same tumor can in principle be exploited to study the spatial distribution of subclones within a tumor mass. However, since the technology required to generate large scDNA sequencing datasets is relatively recent, dedicated analytical approaches are still lacking. RESULTS: We present PhyliCS, the first tool which exploits scCNA data from multiple samples from the same tumor to estimate whether the different clones of a tumor are well mixed or spatially separated. Starting from the CNA data produced with third party instruments, it computes a score, the Spatial Heterogeneity score, aimed at distinguishing spatially intermixed cell populations from spatially segregated ones. Additionally, it provides functionalities to facilitate scDNA analysis, such as feature selection and dimensionality reduction methods, visualization tools and a flexible clustering module. CONCLUSIONS: PhyliCS represents a valuable instrument to explore the extent of spatial heterogeneity in multi-regional tumour sampling, exploiting the potential of scCNA data.

One of the key targets of Industry 4.0 and digital production, in general, is the support of faster, cleaner, and increasingly customizable manufacturing processes. Additive manufacturing (AM) is a natural fit in this context, as it offers the possibility to produce complex parts without the design constraints of traditional manufacturing routes, typically reducing both material waste and time to market. Nonetheless, the lack of repeatability of the manufacturing process, which typically translates into a lack of reproducibility and reliability of the quality of the final products compared to traditional subtractive technologies, is currently one of the major barriers to the widespread adoption of AM in mass production. To overcome this limitation, there are growing efforts in recent years toward better integration of advanced information technologies into AM, exploiting the layer-by-layer nature of the build. The consequence of these efforts is twofold: 1) the integration of advanced sensing technologies into the AM systems, making possible the in situ monitoring of huge amounts of data at multiple time scales and resolutions and 2) the ever-increasing role of data-driven approaches [especially machine learning (ML)] in the analysis of such data to provide real-time quality monitoring and process optimization. This article introduces and reviews the key technological developments of this phenomenon, with a special focus on metal powder bed fusion (PBF) technologies that are attracting the highest attention by the industrial AM community. After introducing the main manufacturing quality issues and needs that have to be developed and optimized, we provide a wide overview of the latest progress of in situ monitoring and control in metal PBF, with special regards to sensing technologies and ML approaches. Finally, we identify the open challenges and future research directions in this field.

MOTIVATION: MicroRNAs (miRNAs) are small RNA molecules (∼22 nucleotide long) involved in post-transcriptional gene regulation. Advances in high-throughput sequencing technologies led to the discovery of isomiRs, which are miRNA sequence variants. While many miRNA-seq analysis tools exist, the diversity of output formats hinders accurate comparisons between tools and precludes data sharing and the development of common downstream analysis methods. RESULTS: To overcome this situation, we present here a community-based project, miRNA Transcriptomic Open Project (miRTOP) working towards the optimization of miRNA analyses. The aim of miRTOP is to promote the development of downstream isomiR analysis tools that are compatible with existing detection and quantification tools. Based on the existing GFF3 format, we first created a new standard format, mirGFF3, for the output of miRNA/isomiR detection and quantification results from small RNA-seq data. Additionally, we developed a command line Python tool, mirtop, to create and manage the mirGFF3 format. Currently, mirtop can convert into mirGFF3 the outputs of commonly used pipelines, such as seqbuster, isomiR-SEA, sRNAbench, Prost! as well as BAM files. Some tools have also incorporated the mirGFF3 format directly into their code, such as, miRge2.0, IsoMIRmap and OptimiR. Its open architecture enables any tool or pipeline to output or convert results into mirGFF3. Collectively, this isomiR categorization system, along with the accompanying mirGFF3 and mirtop API, provide a comprehensive solution for the standardization of miRNA and isomiR annotation, enabling data sharing, reporting, comparative analyses and benchmarking, while promoting the development of common miRNA methods focusing on downstream steps of miRNA detection, annotation and quantification. AVAILABILITY AND IMPLEMENTATION: https://github.com/miRTop/mirGFF3/ and https://github.com/miRTop/mirtop. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

SUMMARY: In the last decade, increasing attention has been paid to the study of gene fusions. However, the problem of determining whether a gene fusion is a cancer driver or just a passenger mutation is still an open issue. Here we present DEEPrior, an inherently flexible deep learning tool with two modes (Inference and Retraining). Inference mode predicts the probability of a gene fusion being involved in an oncogenic process, by directly exploiting the amino acid sequence of the fused protein. Retraining mode allows to obtain a custom prediction model including new data provided by the user. AVAILABILITY AND IMPLEMENTATION: Both DEEPrior and the protein fusions dataset are freely available from GitHub at (https://github.com/bioinformatics-polito/DEEPrior). The tool was designed to operate in Python 3.7, with minimal additional libraries. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: High-throughput next-generation sequencing can generate huge sequence files, whose analysis requires alignment algorithms that are typically very demanding in terms of memory and computational resources. This is a significant issue, especially for machines with limited hardware capabilities. As the redundancy of the sequences typically increases with coverage, collapsing such files into compact sets of non-redundant reads has the 2-fold advantage of reducing file size and speeding-up the alignment, avoiding to map the same sequence multiple times. METHOD: BioSeqZip generates compact and sorted lists of alignment-ready non-redundant sequences, keeping track of their occurrences in the raw files as well as of their quality score information. By exploiting a memory-constrained external sorting algorithm, it can be executed on either single- or multi-sample datasets even on computers with medium computational capabilities. On request, it can even re-expand the compacted files to their original state. RESULTS: Our extensive experiments on RNA-Seq data show that BioSeqZip considerably brings down the computational costs of a standard sequence analysis pipeline, with particular benefits for the alignment procedures that typically have the highest requirements in terms of memory and execution time. In our tests, BioSeqZip was able to compact 2.7 billion of reads into 963 million of unique tags reducing the size of sequence files up to 70% and speeding-up the alignment by 50% at least. AVAILABILITY AND IMPLEMENTATION: BioSeqZip is available at https://github.com/bioinformatics-polito/BioSeqZip. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Latest Next Generation Sequencing technologies opened the way to a novel era of genomic studies, allowing to gain novel insights into multifactorial pathologies as cancer. In particular gene fusion detection and comprehension have been deeply enhanced by these methods. However, state of the art algorithms for gene fusion identification are still challenging. Indeed, they identify huge amounts of poorly overlapping candidates and all the reported fusions should be considered for in lab validation clearly overwhelming wet lab capabilities. RESULTS: In this work we propose a novel methodological approach and tool named FuGePrior for the prioritization of gene fusions from paired-end RNA-Seq data. The proposed pipeline combines state of the art tools for chimeric transcript discovery and prioritization, a series of filtering and processing steps designed by considering modern literature on gene fusions and an analysis on functional reliability of gene fusion structure. CONCLUSIONS: FuGePrior performance has been assessed on two publicly available paired-end RNA-Seq datasets: The first by Edgren and colleagues includes four breast cancer cell lines and a normal breast sample, whereas the second by Ren and colleagues comprises fourteen primary prostate cancer samples and their paired normal counterparts. FuGePrior results accounted for a reduction in the number of fusions output of chimeric transcript discovery tools that ranges from 65 to 75% depending on the considered breast cancer cell line and from 37 to 65% according to the prostate cancer sample under examination. Furthermore, since both datasets come with a partial validation we were able to assess the performance of FuGePrior in correctly prioritizing real gene fusions. Specifically, 25 out of 26 validated fusions in breast cancer dataset have been correctly labelled as reliable and biologically significant. Similarly, 2 out of 5 validated fusions in prostate dataset have been recognized as priority by FuGePrior tool.

BACKGROUND: Massive parallel sequencing of transcriptomes, revealed the presence of many miRNAs and miRNAs variants named isomiRs with a potential role in several cellular processes through their interaction with a target mRNA. Many methods and tools have been recently devised to detect and quantify miRNAs from sequencing data. However, all of them are implemented on top of general purpose alignment methods, thus providing poorly accurate results and no information concerning isomiRs and conserved miRNA-mRNA interaction sites. RESULTS: To overcome these limitations we present a novel algorithm named isomiR-SEA, that is able to provide users with very accurate miRNAs expression levels and both isomiRs and miRNA-mRNA interaction sites precise classifications. Tags are mapped on the known miRNAs sequences thanks to a specialized alignment algorithm developed on top of biological evidence concerning miRNAs structure. Specifically, isomiR-SEA checks for miRNA seed presence in the input tags and evaluates, during all the alignment phases, the positions of the encountered mismatches, thus allowing to distinguish among the different isomiRs and conserved miRNA-mRNA interaction sites. CONCLUSIONS: isomiR-SEA performances have been assessed on two public RNA-Seq datasets proving that the implemented algorithm is able to account for more reliable and accurate miRNAs expression levels with respect to those provided by two compared state of the art tools. Moreover, differently from the few methods currently available to perform isomiRs detection, the proposed algorithm implements the evaluation of isomiRs and conserved miRNA-mRNA interaction sites already in the first alignment phases, thus avoiding any additional filtering stages potentially responsible for the loss of useful information.

Determining the correlation between biomedical terms is a powerful instrument to help scientist research activity, both to understand experimental results and to design new ones. In particular, a great potential comes from the integration of the many heterogeneous information sources currently available on the Web. In this article we focus on the correlation between genes and biological processes. In this context, we present a methodology for integrating information from biomedical literature with other heterogeneous types of structured information. In particular, the information sources integrated in this work are PubMed abstracts, pathway databases, and NCI thesaurus definitions. The integration is performed at the semantic analysis level using a customized approach we developed to modulate the impact of the different sources on the correlation score. We report the results of a study concerning the impact of the information integration on the correlation score and of the user-level parameters we introduced to modulate the impact of pathway data or NCI definitions with respect to biomedical literature information, depending on the context of the search. To evaluate the methodology, we performed correlation measures on six biological processes and nine genes by comparing the results with and without the integration of pathways and NCI definitions.

An effective knowledge extraction and quantification methodology from biomedical literature would allow the researcher to organize and analyze the results of high-throughput experiments on microarrays and next-generation sequencing technologies. Despite the large amount of raw information available on the web, a tool able to extract a measure of the correlation between a list of genes and biological processes is not yet available. In this paper, we present Gelsius, a workflow that incorporates biomedical literature to quantify the correlation between genes and terms describing biological processes. To achieve this target, we build different modules focusing on query expansion and document cononicalization. In this way, we reached to improve the measurement of correlation, performed using a latent semantic analysis approach. To the best of our knowledge, this is the first complete tool able to extract a measure of genes-biological processes correlation from literature. We demonstrate the effectiveness of the proposed workflow on six biological processes and a set of genes, by showing that correlation results for known relationships are in accordance with definitions of gene functions provided by NCI Thesaurus. On the other side, the tool is able to propose new candidate relationships for later experimental validation. The tool is available at >http://bioeda1.polito.it:8080/medSearchServlet/.

MOTIVATION: Next-generation sequencing technology allows the detection of genomic structural variations, novel genes and transcript isoforms from the analysis of high-throughput data. In this work, we propose a new framework for the detection of fusion transcripts through short paired-end reads which integrates splicing-driven alignment and abundance estimation analysis, producing a more accurate set of reads supporting the junction discovery and taking into account also not annotated transcripts. Bellerophontes performs a selection of putative junctions on the basis of a match to an accurate gene fusion model. RESULTS: We report the fusion genes discovered by the proposed framework on experimentally validated biological samples of chronic myelogenous leukemia (CML) and on public NCBI datasets, for which Bellerophontes is able to detect the exact junction sequence. With respect to state-of-art approaches, Bellerophontes detects the same experimentally validated fusions, however, it is more selective on the total number of detected fusions and provides a more accurate set of spanning reads supporting the junctions. We finally report the fusions involving non-annotated transcripts found in CML samples. AVAILABILITY AND IMPLEMENTATION: Bellerophontes JAVA/Perl/Bash software implementation is free and available at http://eda.polito.it/bellerophontes/.

Arterial Spin Labelling (ASL) is a perfusion MRI technique with tremendous applications in the study of biological markers and prognostic factors of brain tumors and in the assessment of neural diseases, moreover, it is completely non-invasive as it uses the magnetically inverted blood of the patient as an endogenous tracer. Unfortunately this powerful method is only viable in very limited conditions due to its extreme sensitivity to artifacts originated by head motion, that are not effectively addressed by the current software solutions. This paper presents a motion correction procedure that addresses this issue and provides improved solutions to enhance ASL images of the brain in presence of severe head motion. Experimental results run on a motion-affected pCASL dataset show the concept and demonstrate the superiority of our proposed procedure compared to standard 3D registration.

One of the most important challenges facing the entire globe is the trend towards an aging population. By 2045, there will be more people over 60 years old than younger than 15, thus raising from 600mln to 2bln worldwide. This will raise the number of patients with age-specific, chronic and degenerative diseases (e.g. cardio-vascular, cancer, diabetes, Alzheimer's, Parkinson's). Minimally-invasive imaging technologies such as PET (Positron Emission Tomography) and MPI (Magnetic Resonance Imaging) play a vital role in detecting and tracking the evolution of the above mentioned illnesses and determining the strategy and the effectiveness of the prescribed therapies. So far the detection unit of PET equipment has been implemented using photomultipliers tubes (PMTs). A novel solid state photo-detector, the Silicon photomultiplier (SiPM), can replace the PMT, offering, among many other advantages, the possibility of PET/MRI combo equipment.

BACKGROUND: Computational methods for microRNA target prediction are a fundamental step to understand the miRNA role in gene regulation, a key process in molecular biology. In this paper we present miREE, a novel microRNA target prediction tool. miREE is an ensemble of two parts entailing complementary but integrated roles in the prediction. The Ab-Initio module leverages upon a genetic algorithmic approach to generate a set of candidate sites on the basis of their microRNA-mRNA duplex stability properties. Then, a Support Vector Machine (SVM) learning module evaluates the impact of microRNA recognition elements on the target gene. As a result the prediction takes into account information regarding both miRNA-target structural stability and accessibility. RESULTS: The proposed method significantly improves the state-of-the-art prediction tools in terms of accuracy with a better balance between specificity and sensitivity, as demonstrated by the experiments conducted on several large datasets across different species. miREE achieves this result by tackling two of the main challenges of current prediction tools: (1) The reduced number of false positives for the Ab-Initio part thanks to the integration of a machine learning module (2) the specificity of the machine learning part, obtained through an innovative technique for rich and representative negative records generation. The validation was conducted on experimental datasets where the miRNA:mRNA interactions had been obtained through (1) direct validation where even the binding site is provided, or through (2) indirect validation, based on gene expression variations obtained from high-throughput experiments where the specific interaction is not validated in detail and consequently the specific binding site is not provided. CONCLUSIONS: The coupling of two parts: a sensitive Ab-Initio module and a selective machine learning part capable of recognizing the false positives, leads to an improved balance between sensitivity and specificity. miREE obtains a reasonable trade-off between filtering false positives and identifying targets. miREE tool is available online at http://didattica-online.polito.it/eda/miREE/

The considerable improvement in the biotechnogical field and the adoption of screen techniques such as high throughput arrays produced a spread of biological and genetical data and scientific papers, mostly diffused on the Web. Even if the huge amount of available information represents a major step forward for the biomedical research field, the main effort for a scientist is to evaluate the correlation among the concepts both in qualitative and in quantitative terms. In the presented work, exploiting the richness of the UMLS Metathesaurus in combination with the novelty of the literature provided by PubMed, an automatic flow aimed at the scoring the semantic correlation among biomedical terms (e. g. biomolecules and biological processes) is proposed. The experiments show that obtained correlations are fully coherent with the information coming from the biological literature. The accuracy of the results remarks the importance of combining text mining techniques with complete and well structured thesaurus, such as UMLS.

Gene regulation is one of the most important processes in the molecular biology, in the last years the microRNA molecule, one of the non-coding RNAs involved in the process, has been the focus of attention for several studies. The computational research on this area has gained a notable importance, considering the low amount of experimental information available and the lack of understanding of the microRNA binding mechanism. This article deals with the microRNA-target prediction and presents an innovative method for it. First it generates a set of promising binding sites for a given microRNA using a Genetic Algorithm, at the same time a set of target genes is selected based on the biological process under study. Secondly the set of promising binding sites is mapped into the selected set of target genes, in order to provide real binding sites and finally the resulting targets are filtered according to a biological or structural property. The objectives are to provide a flexible method that is capable of incorporating easily new knowledge, is independent of availability of the experimental information and is able to give hints on the research towards new characteristics among the microRNA binding sites such as motifs. The results present some of this novel properties and present a comparison with the most frequently used methods in the field.

In this paper we present a fully automated morphology-based technique for segmentation of nuclei in cancer tissue images and we compare it with a common technique for biomedical image processing, namely active contours. We discuss the limitations of active contours in the processing of immunohistochemical images characterized by heterogeneously stained nuclear region and noise caused by the presence of multiple tissue layers in the sample. We describe the integration of the proposed approach in a fully automated protein activity quantification tool. Finally, we demonstrate and motivate through extensive experiments that our fully automated morphology-based approach provides better accuracy compared to various active contours implementations.

When analyzing the relationship between genes under different scenarios, the integration of different microarray experiments becomes a relevant task. This paper presents a framework to address some intrinsic problems of integration, due for instance to scaling issues, error bias, different experimental conditions or technology and protocols. Our approach projects original microarray data in a common transformed space to create a common representation of different microarray datasets. This approach allows us to integrate data from various microarray platforms or microarrays based on different experimental conditions. We validate our framework with experiments on real microarray datasets. The results suggest that our approach can be a profitably exploited for microarray data integration and further gene expression analysis applications.

An accurate and standardized technique for tumor tissue segmentation is a critical step for monitoring and quantifying the activity of specific families of pro- teins involved in multi-factorial genetic pathologies. However, fully automated tissue and cell segmentation in clinical images presents many challenges related to the characteristics of the images that make traditional approaches substantially ineffective or incomplete. In this paper we present a fully-automated algorithm that is able to perform accurate and fast segmentation of tissue images. Experimental results on several real-life datasets demonstrate the high level of accuracy achievable thanks to our approach.

This work presents the first fully-automated computer-aided analysis approach to the quantification of the expression of receptors for the non-small cell lung carcinoma. This immunohistochemical analysis is usually performed by pathologists via visual inspection of tissue samples images. Our techniques streamlines this error-prone and time-consuming process, thereby facilitating analysis and diagnosis. Experimental results on several real-life datasets demonstrate the high quantitative precision of our approach

This paper presents an automated algorithm to determine DNA fragment size from atomic force microscope images and to extract the molecular profiles. The sizing of DNA fragments is a widely used procedure for investigating the physical properties of individual or protein-bound DNA molecules. Several atomic force microscope (AFM) real and computer-generated images were tested for different pixel and fragment sizes and for different background noises. The automated approach minimizes processing time with respect to manual and semi-automated DNA sizing. Moreover, the DNA molecule profile recognition can be used to perform further structural analysis. For computer-generated images, the root mean square error incurred by the automated algorithm in the length estimation is 0.6% for a 7.8 nm image pixel size and 0.34% for a 3.9 nm image pixel size. For AFM real images we obtain a distribution of lengths with a standard deviation of 2.3% of mean and a measured average length very close to the real one, with an error around 0.33%.

An automated algorithm is presented to determine the DNA molecule intrinsic curvature profiles and the molecular spatial orientations in atomic force microscope images. The curvature is composed by static and dynamic contributions. The former is the intrinsic curvature, a function of the DNA nucleotide sequence, while the latter is due to thermal fluctuations. This algorithm allows to reconstruct the intrinsic curvature profile excluding the thermal contribution. The automated algorithm reconstructs the intrinsic curvature profile with a mean square error of 3.812/spl middot/ 10/sup -4/ rads over a profile with a central peak value of 0.196 rads, and 6.1/spl middot/10/sup -3/ rads over a curvature profile with two symmetric peaks of about 0.08 rads. Moreover, it correctly detects the location of the peaks in the molecules with a deviation of about 1% of molecule length.