Biology subjects

Ofer, D.

Publications and source records attributed to Ofer, D..

9 recordsLinked to original sources

bioRxiv2026

DNT: Diploid Genomic Foundation Model

Clinical interpretation of genetic variation depends on the diploid genotype, including zygosity, allele dosage and whether multiple variants occur in cis on the same homologue or in trans on different homologues. Most genomic language models process haploid sequences or combine independently encoded haplotypes downstream, so they do not directly represent the paired genotype in a single sequence. We introduce a reference-aligned diploid encoding for single-nucleotide variants (SNVs) and short insertions and deletions (indels), together with unphased and phase-retaining tokenizers that accept phased genotypes and convert them to single-sequence diploid representation. Using Nucleotide Transformer v3 backbones, we continue training 8-million- and 100-million-parameter models and evaluate an auxiliary Contrastive Phase Loss (CPL) designed to retain the phasing information of the variants in contextual representations. We evaluate on a novel compound-heterozygous benchmark containing 9,460 examples. Models whose inputs did not distinguish relative phase remained near chance, whereas our diploidic models improved discrimination with AUROC 0.649, compared to 0.506 for the vocabulary-adapted control. These findings establish a method for making diploid genotype information accessible to genomic language models, rather than a universal improvement in variant prediction; validation in naturally observed, accurately phased clinical cohorts remains necessary.

bioinformatics↗

bioRxiv2026

MAGI: Mechanistic Consequences of Genetic Variants via Genomic Foundation Models

Clinical variant interpretation requires mechanism-aware evidence to guide diagnosis and clarify the biological consequences of mutations. However, existing computational predictors and genomic foundation models largely function as black boxes, providing pathogenicity labels with limited mechanistic insight or clinical actionability. Here, we present MAGI (Mechanistic Annotation of Genomic Impacts), a novel method that bridges this interpretability gap by unifying clinically relevant variant interpretation with mechanistic genomic analysis. MAGI pipeline leverages a genomic transformer model to quantify the effects of DNA variants across 3,623 functional tracks, encompassing regulatory features, multi-omics datasets, including tissue specificity and chromatin states, and 21 additional molecular annotations of genes and transcripts. These signals are integrated through a deterministic logic layer that maps single-nucleotide variants and indels to explicit molecular consequences. We benchmark MAGI-derived consequences against clinical rationales curated from ClinVar and observe strong concordance that scales with the magnitude of functional disruption. MAGI accurately recapitulates canonical pathogenic mechanisms, including start codon loss, splice site disruption, and regulatory element perturbation, consistent with ClinVar annotations. We further present case studies addressing conflicting or incomplete mechanistic interpretations, as well as variants requiring complex inference. Notably, MAGI is also applicable to non-human genomes and was evaluated on multispecies OMIA pathogenic variants. Collectively, MAGI establishes a generalizable framework that extends beyond clinical diagnostics to enable mechanistic discovery in functional genomics, generating mechanistically grounded, testable hypotheses for variants of uncertain significance (VUS) and variants with discordant clinical interpretations. In several cases, MAGI proposes alternative explanations that challenge existing annotations, providing transparent rationales and experimentally tractable predictions.

genetics↗

bioRxiv2026

Improving Causal Gene Identification Using Large Language Models

Genome-Wide Association Studies (GWAS) have successfully identified numerous loci associated with complex traits and diseases, yet pinpointing causal genes remains a significant challenge. The reliance on simple proximity-based heuristics is often insufficient due to linkage disequilibrium, gene interactions, and regulatory effects. Recent advancements in Large Language Models (LLMs) have demonstrated potential in automating causal gene identification, but their effectiveness remains limited by knowledge representation and retrieval mechanisms. This study builds on previous research by evaluating LLMs for causal gene identification, with a focus on enhancing performance through Retrieval-Augmented Generation (RAG) and the incorporation of genomic distance information. We replicate prior results using smaller model Qwen2.5--assessing their predictive accuracy using a benchmark dataset from Open Targets. We improved the preformences when integrating RAG-based literature retrieval (F1 = 0.795) and gene distance information (F1 = 0.806). However, the combined approach yielded diminishing returns, suggesting interactions between these enhancements. Error analysis revealed that genomic distance features improved predictions by reinforcing established heuristics, while RAG enhanced domain knowledge but occasionally led to semantic biases. These findings highlight the potential of hybrid approaches in leveraging both structured genomic features and unstructured textual data.

bioinformatics↗

bioRxiv2026

GFMBench-API: A Standardized Interface for Benchmarking Genomic Foundation Models

The rapid scaling of Genomic Foundation Models (GFMs) has created a critical need for standardized evaluation frameworks. Current benchmarking practices are often fragmented, relying on model-specific preprocessing and inconsistent metric implementations that hinder reproducible comparisons. We present GFMBench-API, a high-level Python interface designed to unify the evaluation lifecycle of GFMs. GFMBench-API provides a modular "middleware" architecture that decouples model-specific tokenization and embedding logic from task-specific data streams and performance metrics. By standardizing the input/output schemas for common genomic tasks, such as regulatory element prediction, variant effect scoring, and long-range interaction mapping, GFMBench-API enables researchers to integrate new models or tasks with minimal "glue code." Our interface ensures mathematical consistency across evaluations, providing a robust foundation for the transparent and systematic benchmarking of GFMs.

genomics↗

bioRxiv2025

OTRec: Deep learning recommender for prospective druggable disease-target associations

Identifying druggable disease--target associations remains a central challenge in translational medicine, limiting therapeutic discovery and repurposing. Here, we present OTRec, a deep learning--based recommender system that ranks such associations at scale and evaluates them in a temporal hold-out setting. Unlike approaches that rely on manually curated or aggregated evidence scores, OTRec employs a two-tower architecture to learn latent representations from 663,351 disease--target pairs. The model integrates heterogeneous inputs, including textual descriptions, ontology-derived features, and biological annotations such as tractability, Gene Ontology (GO) terms, and pathway information. We perform temporal validation by training on the 2022 Open Targets (OT) release and evaluating on clinical trial data from 2025. OTRec improves on the retrospective OT association score (ROC-AUC: 0.872 {+/-} 0.005 vs 0.559; PR-AUC: 0.288 {+/-} 0.009 vs. 0.08). In 5x5 target-disjoint cross-validation, OTRec reaches ROC-AUC 0.950 and PR-AUC 0.844) improving on the OT evidence score (ROC-AUC 0.91; PR-AUC 0.45). We rank the druggable genome across ~19,000 OT platform (OTP) diseases and release ~282,500 candidate associations above a 0.65 score threshold (in-distribution CV precision 0.92), covering 4,346 diseases including 2,322 orphan diseases, through an interactive prediction platform.

bioinformatics↗

bioRxiv2024

Protein Language Models Expose Viral Mimicryand Immune Escape

MotivationViruses elude the immune system through molecular mimicry, adopting biophysical characteristics of their host. We adapt protein language models (PLMs) to differentiate between human and viral proteins. Understanding where the immune system and our models make mistakes could reveal viral immune escape mechanisms. ResultsWe applied pretrained deep-learning PLMs to predict viral from human proteins. Our predictors show state-of-the-art results with AUC of 99.7%. We use interpretable error analysis models to characterize viral escapers. Altogether, mistakes account for 3.9% of the sequences with viral proteins being disproportionally misclassified. Analysis of external variables, including taxonomy and functional annotations, indicated that errors typically involve proteins with low immunogenic potential, viruses specific to human hosts, and those using reverse-transcriptase enzymes for their replication. Viral families causing chronic infections and immune evasion are further enriched and their protein mimicry potential is discussed. We provide insights into viral adaptation strategies and highlight the combined potential of PLMs and explainable AI in uncovering mechanisms of viral immune escape, contributing to vaccine design and antiviral research. Availability and implementationData and results available in https://github.com/ddofer/ProteinHumVir. Contactmichall@cc.huji.ac.il

bioinformatics↗

bioRxiv2023

Detecting Anomalous Proteins Using Deep Representations

Many advances in biomedicine can be attributed to identifying unusual proteins and genes. Many of these proteins unique properties were discovered by manual inspection, which is becoming infeasible at the scale of modern protein datasets. Here, we propose to tackle this challenge using anomaly detection methods that automatically identify unexpected properties. We adopt a state-of-the-art anomaly detection paradigm from computer vision, to highlight unusual proteins. We generate meaningful representations without labeled inputs, using pretrained deep neural network models. We apply these protein language models (pLM) to detect anomalies in function, phylogenetic families, and segmentation tasks. We compute protein anomaly scores to highlight human prion-like proteins, distinguish viral proteins from their host proteome, and mark non-classical ion/metal binding proteins and enzymes. Other tasks concern segmentation of protein sequences into folded and unstructured regions. We provide candidates for rare functionality (e.g., prion proteins). Additionally, we show the anomaly score is useful in 3D folding-related segmentation. Our novel method shows improved performance over strong baselines and has objectively high performance across a variety of tasks. We conclude that the combination of pLM and anomaly detection techniques is a valid method for discovering a range of global and local protein characteristics.

bioinformatics↗

bioRxiv2022

Inferring microRNA regulation: A proteome perspective

Post-transcriptional regulation in multicellular organisms is mediated by microRNAs. However, the mechanisms that determine if a gene is regulated by miRNAs are poorly understood. Previous works focused mostly on miRNA seed matches and other features of the 3-UTR of transcripts. These common approaches relied on knowledge of the miRNA families, and computational approaches still yield poor, inconsistent results, with many false positives. In this work, we present a different paradigm for predicting miRNA-regulated genes based on proteins. In a novel, automated machine learning framework, we use sequence as well as diverse functional annotations to train models on multiple organisms using experimentally validated data. We present insights from tens of millions of features extracted and ranked from different modalities. We show high predictive performance per organism and in generalization across species. We provide a list of novel predictions for Danio rerio (zebrafish) and Arabidopsis thaliana (mouse-ear cress). We compare genomic models, and observe that our protein model outperforms genomics, while a unified model improves on both. While most membranous and disease related proteins are regulated by miRNAs, we observe the G-protein coupled receptor (GPCR) family is an exception, being mostly unregulated by miRNAs, and we raise possible explanations for this. We further show that the evolutionary conservation among duplicated genes does not imply a coherence in miRNA regulation. We conclude that duplicated genes diverge in their tendency to be miRNA regulated. However, protein function is informative across species in predicting post-transcriptional miRNA regulation in living cells.

bioinformatics↗

bioRxiv2021

ProteinBERT: A universal deep-learning model of protein sequence and function

Self-supervised deep language modeling has shown unprecedented success across natural language tasks, and has recently been repurposed to biological sequences. However, existing models and pretraining methods are designed and optimized for text analysis. We introduce ProteinBERT, a deep language model specifically designed for proteins. Our pretraining scheme consists of masked language modeling combined with a novel task of Gene Ontology (GO) annotation prediction. We introduce novel architectural elements that make the model highly efficient and flexible to very large sequence lengths. The architecture of ProteinBERT consists of both local and global representations, allowing end-to-end processing of these types of inputs and outputs. ProteinBERT obtains state-of-the-art performance on multiple benchmarks covering diverse protein properties (including protein structure, post translational modifications and biophysical attributes), despite using a far smaller model than competing deep-learning methods. Overall, ProteinBERT provides an efficient framework for rapidly training protein predictors, even with limited labeled data. Code and pretrained model weights are available at https://github.com/nadavbra/protein_bert.

bioinformatics↗