Biology subjects

Montazeri, H.

Publications and source records attributed to Montazeri, H..

11 recordsLinked to original sources

bioRxiv2026

Cancer Driver Gene Discovery: A Patient-Level Statistical Framework

Tumor genomes harbor a mixture of neutral and positively selected mutations, yet distinguishing true cancer drivers remains a major challenge. Several factors can obscure the detection of selection signals, among which patient-specific variation in mutational burden plays a significant role. Current approaches often fail to account for the heterogeneity in mutation burden across different patients; in particular, no existing method explicitly accounts for it when integrating both mutation recurrence and functional impact. Here we present iDriver, a probabilistic graphical model that integrates both mutation recurrence and functional impact at the individual-patient level, enabling an enhanced estimation of positive selection across functional genomic elements. Applying iDriver to 29 cancer types, we identify both known and previously unrecognized drivers spanning coding and noncoding regions, and provide evidence for their clinical and biological relevance. In comprehensive benchmarks against 12 established driver discovery methods, iDriver consistently outperformed all competitors, achieving the highest rankings for known cancer drivers across both coding and noncoding elements.

bioinformatics↗

bioRxiv2026

Reliable Evaluation and Learning in Multi-input Biological Association Prediction

Multi-input association prediction is central to many key problems in computational biology, spanning tasks from drug-target and protein-protein interactions to higher-order challenges such as drug synergy modeling and MHC-peptide-TCR binding. Yet widely used benchmarks often overestimate performance by enabling models to exploit degree ratio shortcuts, while alternative out-of-distribution splits are overly restrictive and impractical. Here we introduce an entity-balanced evaluation framework that systematically neutralizes shortcut signals by balancing positive and negative associations at the entity level. This enables fairer assessments that reflect genuine relational learning and extend naturally from pairwise to multi-entity problems. We further present UnbiasNet, a model-agnostic training strategy that cycles through diverse entity-balanced sub-training sets, removing access to degree ratio bias and enhancing robustness. Applied to drug-target interaction and drug synergy prediction, our framework reveals the extent of shortcut reliance in existing methods while enabling consistent identification of meaningful biological associations, thereby setting a rigorous foundation for future methodological progress.

bioinformatics↗

bioRxiv2026

Evaluating Single-Cell Perturbation Response Models Is Far from Straightforward

Predicting cellular responses to genetic and chemical perturbations remains a central challenge in single-cell biology and a key step toward building in silico virtual cells. The rapid growth of perturbation datasets and advances in deep-learning models have raised expectations for accurate and generalizable prediction. We show that these expectations are overly optimistic, largely due to the failure modes of existing evaluation metrics. In this study, using cross-splitting, controlled noise experiments, and synthetic data, we systematically evaluate both prediction models and evaluation metrics. We demonstrate that widely used metrics, including correlation-based measures and common distributional distances, are strongly influenced by scale, sparsity, and dimensionality, often misrepresenting model performance. In particular, the Wasserstein distance fails in high-dimensional gene expression spaces under variance scaling, while the Energy distance can overlook disruptions in gene-gene dependencies. Our analyses further reveal that complex deep learning models often underperform simple baselines and remain far from empirical performance bounds across multiple chemical perturbation datasets. Together, our framework exposes critical pitfalls, establishes robust evaluation guidelines, and provides a foundation for trustworthy benchmarking toward reliable virtual-cell models.

bioinformatics↗

bioRxiv2025

sCIN: A Contrastive Learning Framework for Single-Cell Multi-omics Data Integration

The rapid advancement of single-cell omics technologies such as scRNA-seq and scATAC-seq has transformed our understanding of cellular heterogeneity and regulatory mechanisms. However, integrating these data types remains challenging due to distributional discrepancies and distinct feature spaces. To address this, we present a novel single-cell Contrastive INtegration framework (sCIN), that integrates different omics modalities into a shared low-dimensional latent space. sCIN uses modality-specific encoders and contrastive learning to generate latent representations for each modality, aligning cells across modalities and removing technology-specific biases. The framework was designed to rigorously prevent data leakage between training and testing, and was extensively evaluated on three real-world paired and unpaired datasets including SHARE-seq, 10X PBMC (10k version), and CITE-seq. Paired datasets refer to multi-omics data generated using technologies capable of capturing different omics features from the same individual cells whereas unpaired datasets are collected from different but similar cell populations within the same tissue, with each modality measured separately. Results on paired and unpaired datasets show that sCIN outperforms state-of-the-art models, including scGLUE, scBridge, sciCAN, Con-AAE, Harmony, and MOFA+, across multiple metrics: Average Silhouette Width (ASW) for clustering quality, Recall@k, Cell type@k, Cell type accuracy, and Median Rank for integration quality. Moreover, sCIN was evaluated on simulated unpaired datasets derived from paired data, demonstrating its ability to leverage available biological information for effective multimodal integration. In summary, sCIN reliably integrates omics modalities while preserving biological meaning in both paired and unpaired settings.

bioinformatics↗

bioRxiv2023

Uncovering Hidden Cancer Self-Dependencies through Analysis of shRNA-Level Dependency Scores

Large-scale short hairpin RNA (shRNA) screens on well-characterized human cancer cell lines have been widely used to identify novel cancer dependencies. However, the off-target effects of shRNA reagents pose a significant challenge in the analysis of these screens. To mitigate these off-target effects, various approaches have been proposed that aggregate different shRNA viability scores targeting a gene into a single gene-level viability score. Most computational methods for discovering cancer dependencies rely on these gene-level scores. In this paper, we propose a computational method, named NBDep, to find cancer self-dependencies by directly analyzing shRNA-level dependency scores instead of gene-level scores. The NBDep algorithm begins by removing known batch effects of the shRNAs and selecting a subset of concordant shRNAs for each gene. It then uses negative binomial random effects models to statistically assess the dependency between genetic alterations and the viabilities of cell lines by incorporating all shRNA dependency scores of each gene into the model. We applied NBDep to the shRNA dependency scores available at Project DRIVE, which covers 26 different types of cancer. The proposed method identified more well-known and putative cancer genes compared to alternative gene-level approaches in pan-cancer and cancer-specific analyses. Additionally, we demonstrated that NBDep controls type-I error and outperforms statistical tests based on gene-level scores in simulation studies. Author SummaryLarge-scale shRNA screening is increasingly being used in cancer genomics to discover genes involved in cancer by analyzing the viabilities of cell lines upon knocking down a gene using a pool of short hairpin RNAs (shRNA). However, off-target effects, which result from the knockdown of unintended genes, are a major issue in RNAi screening. To address this issue, various computational methods have been developed to aggregate shRNA viability scores into gene-level dependency scores. In this paper, we propose a method called NBDep to identify cancer gene drivers that directly addresses the challenge of off-target effects at the shRNA level. NBDep identifies cancer gene drivers in three classes: amplification, missense, and non-missense alterations. In this method, we first remove known batch effects, select a subset of the most consistent shRNAs of each gene, and then perform a negative binomial mixed-effect model. The NBDep method not only identifies well-recognized and novel cancer driver genes but also has more statistical power than gene-level-score methods while controlling type-error. In summary, NBDep presents a new technique for analyzing shRNA screens and has the potential to uncover previously unknown cancer dependencies.

bioinformatics↗

bioRxiv2022

A Systems Biology Approach to Pathogenesis of Gastric Cancer: Gene Network Modeling and Pathway Analysis

IntroductionGastric cancer (GC) ranks among the most common malignancies worldwide. In our previous study, we found overexpressed genes in GC clinical samples. The goal of the current study was to find critical genes and key pathways involved in the pathogenesis of GC. MethodsGene interactions were analyzed using STRING, and Cytoscape was used to visualize the molecular interaction network. CytoHubba was used for drawing the PPI network and identifying hub proteins. The Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) at STRING were used for the enrichment analysis of the hub genes. Cluster analysis of the network was done using CytoCluster. MEME Suite was used for promoter analysis of the hub genes using Tomtom and GoMo tools. Results and DiscussionOur results showed that the most affected processes in GC are the metabolic processes. The OXPHOS pathway was also considerably enriched in our analyses. These results showed the significant role of mitochondria in GC pathogenesis. Although many investigations have focused on the mitochondrial role in the pathogenesis of various cancers, the characteristics of respiratory and metabolic changes in GC have not been fully elucidated. Our results also showed that most of the affected pathways in GC were the pathways also involved in neurodegenerative diseases. Also, promoter analysis showed that negative regulation of signal transduction might play an important role in GC pathogenesis. The results of this study might open up new insights into GC pathogenesis. The identified genes might be novel diagnostic or prognostic biomarkers or potential therapeutic targets for GC. Nonetheless, these results were obtained by bioinformatics analysis and require further clinical validation.

systems biology↗

bioRxiv2022

The Functional Landscape of SARS-CoV-2 3CL Protease

SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) as the etiologic agent of COVID-19 (coronavirus disease 2019) has drastically altered life globally. Numerous efforts have been placed on the development of therapeutics to treat SARS-CoV-2 infection. One particular target is the 3CL protease (3CLpro), which holds promise as it is essential to the virus and highly conserved among coronaviruses, suggesting that it may be possible to find broad inhibitors that treat not just SARS-CoV-2 but other coronavirus infections as well. While the 3CL protease has been studied by many groups for SARS-CoV-2 and other coronaviruses, our understanding of its tolerance to mutations is limited, knowledge which is particularly important as 3CL protease inhibitors become utilized clinically. Here, we develop a yeast-based deep mutational scanning approach to systematically profile the activity of all possible single mutants of the SARS-CoV-2 3CLpro, and validate our results both in yeast and in authentic viruses. We reveal that the 3CLpro is highly malleable and is capable of tolerating mutations throughout the protein, including within the substrate binding pocket. Yet, we also identify specific residues that appear immutable for function of the protease, suggesting that these interactions may be novel targets for the design of future 3CLpro inhibitors. Finally, we utilize our screening results as a basis to identify E166V as a resistance-conferring mutation against the therapeutic 3CLpro inhibitor, nirmatrelvir, in clinical use. Collectively, the functional map presented herein may serve as a guide for further understanding of the biological properties of the 3CL protease and for drug development for current and future coronavirus pandemics.

microbiology↗

bioRxiv2020

Comparing mutational pathways to lopinavir resistance in HIV-1 subtypes B versus C

Although combination antiretoviral therapies seem to be effective at controlling HIV-1 infections regardless of the viral subtype, there is increasing evidence for subtype-specific drug resistance mutations. The order and rates at which resistance mutations accumulate in different subtypes also remain poorly understood. Here, we present a methodology for the comparison of mutational pathways in different HIV-1 subtypes, based on Hidden Conjunctive Bayesian Networks (H-CBN), a probabilistic model for inferring mutational pathways from cross-sectional genotype data. We introduce a Monte Carlo sampling scheme for learning H-CBN models on a large number of resistance mutations and develop a statistical test to assess differences in the inferred mutational pathways between two groups. We apply this method to the temporal progression of mutations conferring resistance to the protease inhibitor lopinavir in a large cross-sectional data set of South African individuals living with HIV-1 subtype C, as well as a genotype data set of subtype B infections derived from the Stanford HIV Drug Resistance Database and the Swiss HIV Cohort Study. We find strong support for different initial mutational events in the protease, namely at residue 46 in subtype B and at residue 82 in subtype C. Our results also show that mutations can accumulate along various alternative paths within subtypes, as opposed to a unique total temporal ordering. Furthermore, the maximum likelihood mutational networks for subtypes B and C share only 7 edges (Jaccard distance 0.802) and imply many different evolutionary pathways. Beyond HIV drug resistance, the statistical methodology is applicable more generally for the comparison of inferred mutational pathways between any two groups. Author summaryThere is a disparity in the distribution of infections by HIV-1 subtype in the world. Subtype B is predominant in America, Western Europe and Australia, and most therapeutic strategies are based on research and clinical studies on this subtype. However, non-B subtypes represent the majority of global HIV-1 infections; e.g., subtype C alone accounts for nearly half of all HIV-1 infections. We present a statistical framework enabling the comparison of patterns of accumulating mutations in different HIV-1 subtypes. Specifically, we study lopinavir resistance pathways in HIV-1 subtypes B versus C, but the methodology can be generally applied to compare the temporal ordering of genetic events in different subgroups.

evolutionary biology↗

bioRxiv2020

GATA3 and MDM2 are synthetic lethal in estrogen receptor-positive breast cancers

Synthetic lethal interactions, where the simultaneous but not individual inactivation of two genes is lethal to the cell, have been successfully exploited to treat cancer. GATA3 is frequently mutated in estrogen receptor (ER)-positive breast cancers and its deficiency defines a subset of patients with poor response to hormonal therapy and poor prognosis. However, GATA3 is not targetable. Here we show that GATA3 and MDM2 are synthetically lethal in ER-positive breast cancer. Depletion and pharmacological inhibition of MDM2 induce apoptosis in GATA3-deficient models in vitro, in vivo and in patient-derived organoids (PDOs) harboring GATA3 somatic mutation. The synthetic lethality requires p53 and acts via the PI3K/Akt/mTOR pathway. Our results present MDM2 as a novel therapeutic target in the substantial cohort of ER-positive, GATA3-mutant breast cancer patients. With MDM2 inhibitors widely available, our findings can be rapidly translated into clinical trials to evaluate in-patient efficacy.

cancer biology↗

bioRxiv2019

APSiC: Analysis of Perturbation Screens for the Identification of Novel Cancer Genes

BackgroundSystematic perturbation screens provide comprehensive resources for the elucidation of cancer driver genes, including rarely mutated genes that are missed by approaches focused on frequently mutated genes and driver genes for which the basis for oncogenicity is non-genetic. The perturbation of many genes in relatively few cell lines in such functional screens necessitates the development of specialized computational tools with sufficient statistical power. ResultsHere we developed APSiC (Analysis of Perturbation Screens for identifying novel Cancer genes) that can identify genetic and non-genetic drivers even with a limited number of samples. Applying APSiC to the large-scale deep shRNA screen Project DRIVE, APSiC identified well-known, pan-cancer genetic drivers, novel putative genetic drivers known to be dysregulated in specific cancer types and the context dependency of mRNA-splicing between cancer types. Additionally, APSiC discovered a median of 28 and 35 putative non-genetic oncogenes and tumor suppressor genes, respectively, for individual cancer types, including genes involved in genome stability maintenance and cell cycle. We functionally demonstrated that LRRC4B, a putative non-genetic tumor suppressor gene that has not previously been associated with carcinogenesis, suppresses proliferation by delaying cell cycle and modulates apoptosis in breast cancer. ConclusionWe demonstrate APSiC is a robust statistical framework for discovery of novel cancer genes through analysis of large-scale perturbation screens. The analysis of DRIVE using APSiC is provided as a web portal and represents a valuable resource for the discovery of novel cancer genes.

genomics↗

bioRxiv2019

Discovery of synthetic lethal interactions from large-scale pan-cancer perturbation screens

Despite the progress in precision oncology, development of cancer therapies is limited by the dearth of suitable drug targets1. Novel candidate drug targets can be identified based on the concept of synthetic lethality (SL), which refers to pairs of genes for which an aberration in either gene alone is non-lethal, but co-occurrence of the aberrations is lethal to the cell. We developed SLIdR (Synthetic Lethal Identification in R), a statistical framework for identifying SL pairs from large-scale perturbation screens. SLIdR successfully predicts SL pairs even with small sample sizes while minimizing the number of false positive targets. We applied SLIdR to Project DRIVE data2 and found both established and novel pan-cancer and cancer type-specific SL pairs. We identified and experimentally validated a novel SL interaction between AXIN1 and URI1 in hepatocellular carcinoma, thus corroborating the potential of SLIdR to identify new SL-based drug targets.

bioinformatics↗