MOTIVATION: Sequence-to-expression models typically do not consider chromatin accessibility, a major factor limiting gene regulation. We hypothesized that supplying accessibility as an input feature would allow a sequence-to-expression model to focus on important open regions of the genome. RESULTS: We found that the performance of such an augmented model was significantly better than that of sequence-only or accessibility-only models with similar architectures. Specifically, its ability to predict the expression of highly variable genes and gene expression in other cell types improved, and higher attribution scores in the input DNA sequences of the augmented model conformed to accessibility, enabling the learning of cell type-specific sequence patterns. Additionally, we show that fine-tuning a pre-trained sequence-only model with both sequence and accessibility can boost performance further and highlight the importance of sequencing depth in sequence-to-expression prediction. AVAILABILITY AND IMPLEMENTATION: Source code is available on GitHub at https://github.com/lapohosorsolya/accessible_seq2exp.

Deciphering the underlying gene regulatory networks (GRNs) that govern early human embryogenesis is critical for understanding developmental mechanisms yet remains challenging due to limited sample availability and the inherent complexity of the biological processes involved. To address this, we developed InPheRNo-ChIP, a computational framework that integrates multimodal data, including RNA-seq, transcription factor (TF)-specific ChIP-seq, and phenotypic labels, to reconstruct phenotype-relevant GRNs associated with endoderm development. The core of this method is a probabilistic graphical model that models the simultaneous effect of TFs on their putative target genes to influence a particular phenotypic outcome. Unlike the majority of existing GRN inference methods that are agnostic to the phenotypic outcomes, InPheRNo-ChIP directly incorporates phenotypic information during GRN inference, enabling the distinction between lineage-specific and general regulatory interactions. We integrated data from three experimental studies and applied InPheRNo-ChIP to infer the GRN governing the differentiation of human embryonic stem cells into definitive endoderm. Benchmarking against a scRNA-seq CRISPRi study demonstrated InPheRNo-ChIP's ability to identify regulatory interactions involving endoderm markers FOXA2, SMAD2, and SOX17, outperforming other methods. This highlights the importance of incorporating the phenotypic context during network inference. Furthermore, an ablation study confirms the synergistic contribution of ChIP-seq, RNA-seq, and phenotypic data, highlighting the value of multimodal integration for accurate phenotype-relevant GRN reconstruction.

An overwhelming majority of protein-protein interaction (PPI) studies are conducted in a select few model organisms largely due to constraints in time and cost of the associated 'wet lab' experiments. In silico PPI inference methods are ideal tools to overcome these limitations, but often struggle with cross-species predictions. We present INTREPPPID, a method that incorporates orthology data using a new 'quintuplet' neural network, which is constructed with five parallel encoders with shared parameters. INTREPPPID incorporates both a PPI classification task and an orthologous locality task. The latter learns embeddings of orthologues that have small Euclidean distances between them and large distances between embeddings of all other proteins. INTREPPPID outperforms all other leading PPI inference methods tested on both the intraspecies and cross-species tasks using strict evaluation datasets. We show that INTREPPPID's orthologous locality loss increases performance because of the biological relevance of the orthologue data and not due to some other specious aspect of the architecture. Finally, we introduce PPI.bio and PPI Origami, a web server interface for INTREPPPID and a software tool for creating strict evaluation datasets, respectively. Together, these two initiatives aim to make both the use and development of PPI inference tools more accessible to the community.

MOTIVATION: Combination therapies have emerged as a treatment strategy for cancers to reduce the probability of drug resistance and to improve outcomes. Large databases curating the results of many drug screening studies on preclinical cancer cell lines have been developed, capturing the synergistic and antagonistic effects of combination of drugs in different cell lines. However, due to the high cost of drug screening experiments and the sheer size of possible drug combinations, these databases are quite sparse. This necessitates the development of transductive computational models to accurately impute these missing values. RESULTS: Here, we developed MARSY, a deep-learning multitask model that incorporates information on the gene expression profile of cancer cell lines, as well as the differential expression signature induced by each drug to predict drug-pair synergy scores. By utilizing two encoders to capture the interplay between the drug pairs, as well as the drug pairs and cell lines, and by adding auxiliary tasks in the predictor, MARSY learns latent embeddings that improve the prediction performance compared to state-of-the-art and traditional machine-learning models. Using MARSY, we then predicted the synergy scores of 133 722 new drug-pair cell line combinations, which we have made available to the community as part of this study. Moreover, we validated various insights obtained from these novel predictions using independent studies, confirming the ability of MARSY in making accurate novel predictions. AVAILABILITY AND IMPLEMENTATION: An implementation of the algorithms in Python and cleaned input datasets are provided in https://github.com/Emad-COMBINE-lab/MARSY.

MOTIVATION: Interpretable deep learning (DL) models that can provide biological insights, in addition to accurate predictions, are of great interest to the biomedical community. Recently, interpretable DL models that incorporate signaling pathways have been proposed for drug response prediction (DRP). While these models improve interpretability, it is unclear whether this comes at the cost of less accurate DRPs, or a prediction improvement can also be obtained. RESULTS: We comprehensively and systematically assessed four state-of-the-art interpretable DL models using three pathway collections to assess their ability in making accurate predictions on unseen samples from the same dataset, as well as their generalizability to an independent dataset. Our results showed that models that explicitly incorporate pathway information in the form of a latent layer perform worse compared to models that incorporate this information implicitly. However, in most evaluation setups, the best performance was achieved using a black-box multilayer perceptron, and the performance of a random forests baseline was comparable to those of the interpretable models. Replacing the signaling pathways with randomly generated pathways showed a comparable performance for the majority of the models. Finally, the performance of all models deteriorated when applied to an independent dataset. These results highlight the importance of systematic evaluation of newly proposed models using carefully selected baselines. We provide different evaluation setups and baseline models that can be used to achieve this goal. AVAILABILITY AND IMPLEMENTATION: Implemented models and datasets are provided at https://doi.org/10.5281/zenodo.7787178 and https://doi.org/10.5281/zenodo.7101665, respectively.

MOTIVATION: The increasing number of publicly available databases containing drugs' chemical structures, their response in cell lines, and molecular profiles of the cell lines has garnered attention to the problem of drug response prediction. However, many existing methods do not fully leverage the information that is shared among cell lines and drugs with similar structure. As such, drug similarities in terms of cell line responses and chemical structures could prove to be useful in forming drug representations to improve drug response prediction accuracy. RESULTS: We present two deep learning approaches, BiG-DRP and BiG-DRP+, for drug response prediction. Our models take advantage of the drugs' chemical structure and the underlying relationships of drugs and cell lines through a bipartite graph and a heterogeneous graph convolutional network that incorporate sensitive and resistant cell line information in forming drug representations. Evaluation of our methods and other state-of-the-art models in different scenarios shows that incorporating this bipartite graph significantly improves the prediction performance. In addition, genes that contribute significantly to the performance of our models also point to important biological processes and signaling pathways. Analysis of predicted drug response of patients' tumors using our model revealed important associations between mutations and drug sensitivity, illustrating the utility of our model in pharmacogenomics studies. AVAILABILITY AND IMPLEMENTATION: An implementation of the algorithms in Python is provided in https://github.com/ddhostallero/BiG-DRP. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Computational methods for the prediction of protein-protein interactions (PPIs), while important tools for researchers, are plagued by challenges in generalizing to unseen proteins. Datasets used for modelling protein-protein predictions are particularly predisposed to information leakage and sampling biases. RESULTS: In this study, we introduce RAPPPID, a method for the Regularized Automatic Prediction of Protein-Protein Interactions using Deep Learning. RAPPPID is a twin Averaged Weight-Dropped Long Short-Term memory network which employs multiple regularization methods during training time to learn generalized weights. Testing on stringent interaction datasets composed of proteins not seen during training, RAPPPID outperforms state-of-the-art methods. Further experiments show that RAPPPID's performance holds regardless of the particular proteins in the testing set and its performance is higher for experimentally supported edges. This study serves to demonstrate that appropriate regularization is an important component of overcoming the challenges of creating models for PPI prediction that generalize to unseen proteins. Additionally, as part of this study, we provide datasets corresponding to several data splits of various strictness, in order to facilitate assessment of PPI reconstruction methods by others in the future. AVAILABILITY AND IMPLEMENTATION: Code and datasets are freely available at https://github.com/jszym/rapppid and Zenodo.org. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Abnormal coagulation and an increased risk of thrombosis are features of severe COVID-19, with parallels proposed with hemophagocytic lymphohistiocytosis (HLH), a life-threating condition associated with hyperinflammation. The presence of HLH was described in severely ill patients during the H1N1 influenza epidemic, presenting with pulmonary vascular thrombosis. We tested the hypothesis that genes causing primary HLH regulate pathways linking pulmonary thromboembolism to the presence of SARS-CoV-2 using novel network-informed computational algorithms. This approach led to the identification of Neutrophils Extracellular Traps (NETs) as plausible mediators of vascular thrombosis in severe COVID-19 in children and adults. Taken together, the network-informed analysis led us to propose the following model: the release of NETs in response to inflammatory signals acting in concert with SARS-CoV-2 damage the endothelium and direct platelet-activation promoting abnormal coagulation leading to serious complications of COVID-19. The underlying hypothesis is that genetic and/or environmental conditions that favor the release of NETs may predispose individuals to thrombotic complications of COVID-19 due to an increase risk of abnormal coagulation. This would be a common pathogenic mechanism in conditions including autoimmune/infectious diseases, hematologic and metabolic disorders.

Prediction of clinical drug response (CDR) of cancer patients, based on their clinical and molecular profiles obtained prior to administration of the drug, can play a significant role in individualized medicine. Machine learning models have the potential to address this issue but training them requires data from a large number of patients treated with each drug, limiting their feasibility. While large databases of drug response and molecular profiles of preclinical in-vitro cancer cell lines (CCLs) exist for many drugs, it is unclear whether preclinical samples can be used to predict CDR of real patients. We designed a systematic approach to evaluate how well different algorithms, trained on gene expression and drug response of CCLs, can predict CDR of patients. Using data from two large databases, we evaluated various linear and non-linear algorithms, some of which utilized information on gene interactions. Then, we developed a new algorithm called TG-LASSO that explicitly integrates information on samples' tissue of origin with gene expression profiles to improve prediction performance. Our results showed that regularized regression methods provide better prediction performance. However, including the network information or common methods of including information on the tissue of origin did not improve the results. On the other hand, TG-LASSO improved the predictions and distinguished resistant and sensitive patients for 7 out of 13 drugs. Additionally, TG-LASSO identified genes associated with the drug response, including known targets and pathways involved in the drugs' mechanism of action. Moreover, genes identified by TG-LASSO for multiple drugs in a tissue were associated with patient survival. In summary, our analysis suggests that preclinical samples can be used to predict CDR of patients and identify biomarkers of drug sensitivity and survival.

Predicting the response of cancer cells to drugs is an important problem in pharmacogenomics. Recent efforts in generation of large scale datasets profiling gene expression and drug sensitivity in cell lines have provided a unique opportunity to study this problem. However, one major challenge is the small number of samples (cell lines) compared to the number of features (genes) even in these large datasets. We propose a collaborative filtering like algorithm for modeling gene-drug relationship to identify patients most likely to benefit from a treatment. Due to the correlation of gene expressions in different cell lines, the gene expression matrix is approximately low-rank, which suggests that drug responses could be estimated from a reduced dimension latent space of the gene expression. Towards this end, we propose a joint low-rank matrix factorization and latent linear regression approach. Experiments with data from the Genomics of Drug Sensitivity in Cancer database are included to show that the proposed method can predict drug-gene associations better than the state-of-the-art methods.

Motivation: Cancer genomes exhibit a large number of different alterations that affect many genes in a diverse manner. An improved understanding of the generative mechanisms behind the mutation rules and their influence on gene community behavior is of great importance for the study of cancer. Results: To expand our capability to analyze combinatorial patterns of cancer alterations, we developed a rigorous methodology for cancer mutation pattern discovery based on a new, constrained form of correlation clustering. Our new algorithm, named C3 (Cancer Correlation Clustering), leverages mutual exclusivity of mutations, patient coverage and driver network concentration principles. To test C3, we performed a detailed analysis on TCGA breast cancer and glioblastoma data and showed that our algorithm outperforms the state-of-the-art CoMEt method in terms of discovering mutually exclusive gene modules and identifying biologically relevant driver genes. The proposed agnostic clustering method represents a unique tool for efficient and reliable identification of mutation patterns and driver pathways in large-scale cancer genomics studies, and it may also be used for other clustering problems on biological graphs. Availability and Implementation: The source code for the C3 method can be found at https://github.com/jackhou2/C3 Contacts: [email protected] or [email protected] Supplementary information: Supplementary data are available at Bioinformatics online.

We analyze a new group-testing scheme, termed semi-quantitative group testing, which may be viewed as a concatenation of an adder channel and a discrete quantizer. Our focus is on non-uniform quantizers with arbitrary thresholds. For the most general semi-quantitative group-testing model, we define three new families of sequences capturing the constraints on the code design imposed by the choice of the thresholds. The sequences represent extensions and generalizations of Bhand certain types of super-increasing and lexicographically ordered sequences, and they lead to code structures amenable for efficient recursive decoding. We describe the decoding methods and provide an accompanying computational complexity and performance analysis.

We introduce a novel probabilistic group testing framework, termed Poisson group testing, in which the number of defectives follows a right-truncated Poisson distribution. The Poisson model applies to a number of biological testing scenarios, where the subjects are assumed to be ordered based on their arrival times and where the probability of being defective decreases with time. Our main result is an information-theoretic upper bound on the minimum number of tests required to achieve an average probability of detection error asymptotically converging to zero.

We present a new family of codes for non-uniformly quantized adder channels. Quantized adder channels are generalizations of group testing models, which were studied under the name of semi-quantitative group testing. We describe non-binary group testing schemes in which the test matrices are generated by concatenating scaled disjunct codebooks, with the scaling parameters determined through lexicographical ordering constraints. In addition, we propose simple iterative decoding methods for one class of such codes.

We propose a novel group testing method, termed semiquantitative group testing (SQGT), motivated by a class of problems arising in genome screening experiments. The SQGT is a (possibly) nonbinary pooling scheme that may be viewed as a concatenation of an adder channel and an integer-valued quantizer. In its full generality, SQGT may be viewed as a unifying framework for group testing, in the sense that most group testing models are special instances of SQGT. For the new testing scheme, we define the notion of SQ-disjunct and SQ-separable codes, representing generalizations of classical disjunct and separable codes. We describe several combinatorial and probabilistic constructions for such codes. While for most of these constructions, we assume that the number of defectives is much smaller than total number of test subjects, we also consider the case in which there is no restriction on the number of defectives and they may be as large as the total number of subjects. For the codes constructed in this paper, we describe a number of efficient decoding algorithms. In addition, we describe a belief propagation decoder for sparse SQGT codes for which no other efficient decoder is currently known.

We consider a novel approach to the information bottleneck problem where the goal is to perform compression of a noisy signal, while retaining a significant amount of information about a correlated auxiliary signal. To facilitate analysis, we cast compression with side information as an optimization problem involving an information measure, which for jointly Gaussian random variables equals the classical mutual information. We provide closed form expressions for locally optimal linear compression schemes; in particular, we show that the optimal solutions are of the form of the product of an arbitrary full-rank matrix and the left eigenvectors corresponding to smallest eigenvalues of a matrix related to the signals' covariance matrices. In addition, we study the influence of the sparsity level of the Bernoulli-Gaussian noise on the compression rate. We also highlight the similarities and differences between the noisy bottleneck problem and canonical correlation analysis (CCA), as well as the Gaussian information bottleneck problem.

Metagenomics is an emerging field of molecular biology concerned with analyzing the genomes of environmental samples comprising many different diverse organisms. Given the nature of metagenomic data, one usually has to sequence the genomic material of all organisms in a batch, leading to a mix of reads coming from different DNA sequences. In deep high-throughput sequencing experiments, the volume of the raw reads is extremely high, frequently exceeding 600 Gb. With an ever increasing demand for storing such reads for future studies, the issue of efficient metagenomic compression becomes of paramount importance. We present the first known approach to metagenome read compression, termed MCUIUC (Metagenomic Compression at UIUC). The gist of the proposed algorithm is to perform classification of reads based on unique organism identifiers, followed by reference-based alignment of reads for individually identified organisms, and metagenomic assembly of unclassified reads. Once assembly and classification are completed, lossless reference based compression is performed via positional encoding. We evaluate the performance of the algorithm on moderate sized synthetic metagenomic samples involving 15 randomly selected organisms and describe future directions for improving the proposed compression method.

We consider a novel group testing procedure, termed semi-quantitative group testing, motivated by a class of problems arising in genome sequence processing. Semi-quantitative group testing (SQGT) is a non-binary pooling scheme that may be viewed as a combination of an adder model followed by a quantizer. For the new testing scheme we define the capacity and evaluate the capacity for some special choices of parameters using information theoretic methods. We also define a new class of disjunct codes suitable for SQGT, termed SQ-disjunct codes. We also provide both explicit and probabilistic code construction methods for SQGT with simple decoding algorithms.

We consider the problem of noiseless and noisy low- rank tensor completion from a set of random linear measurements. In our derivations, we assume that the entries of the tensor belong to a finite field of arbitrary size and that reconstruction is based on a rank minimization framework. The derived results show that the smallest number of measurements needed for exact reconstruction is upper bounded by the product of the rank, the order, and the dimension of a cubic tensor. Furthermore, this condition is also sufficient for unique minimization. Similar bounds hold for the noisy rank minimization scenario, except for a scaling function that depends on the channel error probability.

We describe a generalization of the group testing problem termed symmetric group testing. Unlike in classical binary group testing, the roles played by the input symbols zero and one are “symmetric” while the outputs are drawn from a ternary alphabet. Using an information-theoretic approach, we derive sufficient and necessary conditions for the number of tests required for noise-free and noisy reconstructions. Furthermore, we extend the notion of disjunct (zero-false-drop) and separable (uniquely decipherable) codes to the case of symmetric group testing. For the new family of codes, we derive bounds on their size based on probabilistic methods, and provide construction methods based on coding theoretic ideas.

Two effects of a single interference on the performance of an automatic frequency control (AFC) loop are explained, and two measures, average switching rate and mean time to loss of lock, are presented to capture the impact of each effect. These measures are derived in closed-form expressions and single integral formulas in dissimilar fading channels with Rayleigh, Rician, and Nakagami-m distributions. The general case of modulated carriers as well as different special cases are considered. Numerical examples are provided to illustrate the effects of different fading scenarios on the performance of an AFC. The effect of maximum Doppler frequency on the performance of the AFC is also investigated.

Bandwidth efficient signaling requires using pulse-shaping that inherently has intersymbol interference (ISI) except when the timing is perfect. Several lower bounds to the mean-square-error (MSE) of non-data-aided bit-synchronizers are derived for this case. The optimal lower bound is derived using the maximum likelihood (ML) criterion for a sequence of binary pulse amplitude modulated pulses in the presence of ISI and Gaussian noise. This lower bound is used as a benchmark to evaluate the performance of other synchronizers in a practical scenario. It is shown that a previous lower bound based on the ISI-free ML synchronizer cannot be used to lower bound the MSE of bit-synchronizers. A detection theory bound (DTB) (also called Ziv-Zakai bound) is applied to the symbol timing recovery problem in the presence of ISI and it is shown that this bound is a tight lower bound on the MSE of the ML synchronizer. A simple lower bound on this DTB is derived and it is shown that the simple bound is almost as tight as the well known modified Cramer-Rao bound (MCRB) at moderate values of SNR, while it does not suffer from the shortcomings of the MCRB at small values of SNR.

The performance of an automatic frequency control (AFC) loop is investigated using two measures, the mean time to loss of lock and the average switching rate. The AFC is considered to operate in fading while a single interferer is also present at its input. Channels are modeled using independent non-identically distributed Rayleigh, Rician and Nakagami-{m} fading while the special case of independent identically distributed channels is also considered. Closed-form expressions and integral form formulas are derived for the general case of modulated signals as well as important special cases of similar modulations and unmodulated signals. The considered scenarios include previous results from the literature as special cases. Corresponding numerical examples are provided and discussed to illustrate the performance of an AFC in such scenarios.

The maximum likelihood (ML) criterion for symbol timing estimation is derived for a sequence of pulse amplitude modulated pulses in the presence of intersymbol interference (ISI) and Gaussian noise. The performance of this synchronizer is used as a benchmark to evaluate the performance of other synchronizers in a practical scenario. It is shown that a previous lower bound based on the ISI-free ML synchronizer cannot be used to lower bound the mean square error (MSE) of bit-synchronizers. A detection theory bound (DTB) is applied to the symbol timing recovery problem in an ISI channel and it is shown that this bound is a tight lower bound on the MSE of the ML synchronizer. A simple lower bound on this DTB is derived and it is shown that the simple bound is almost as tight as the well known modified Cramer-Rao bound (MCRB) at moderate values of SNR while it does not suffer from the shortcomings of the MCRB at small values of SNR.

The performance of an automatic frequency control (AFC) loop in a noisy fading channel when an interference signal is present at the input of the AFC is studied. Independent non-identically distributed (i.n.d.) channels with Rayleigh and Rician fading are considered. The received signals are assumed to be narrowband and linearly modulated while the analysis is applicable to the unmodulated scenario as well. Closed-form expressions and integral form formulas are derived for the mean time to loss of lock (MTLL) and the average switching rate (ASR) of an AFC. Numerical examples are provided to illustrate the effects of noise and slow fading on the performance of an AFC in the presence of an interferer. It is shown that in some scenarios, an AFC has a better performance if the desired signal is corrupted by more noise.

The mean time to loss of lock and the average switching rate of an automatic frequency control (AFC) loop are derived in this paper for the case of two received signals in dissimilar fading channels. The general case of modulated carriers is considered while the results are also applicable to unmodulated carriers. The channels are assumed to have Rayleigh, Rician and Nakagami-m distributions. Numerical examples are provided to illustrate the effect of these fading scenarios on the performance of an AFC in the presence of cochannel interference.

The mean time to loss of lock and the average switching rate of an automatic frequency control loop operating in fading in the presence of a single interferer are derived. Closed-form expressions and integral form formulas are derived for the general case of modulated carriers as well as important special cases of similar modulations and unmodulated carriers. The general results include, as special cases, some previous more restricted results. Fading channels are assumed to be independent non-identically distributed (i.n.d.) with Rayleigh, Rician and Nakagami-m distributions while the special case of independent identically distributed (i.i.d.) channels is also considered. Corresponding numerical examples are provided and discussed to illustrate the results.