arXiv subjects

Michal Linial

Publications and source records attributed to Michal Linial.

5 recordsLinked to original sources

arXiv2026

ProteinJEPA: Latent prediction complements protein language models

Protein language models are trained primarily with masked language modeling (MLM), which predicts amino-acid identities at masked positions. We ask whether latent-space prediction can complement these token-level objectives under matched wall-clock budget. Across pretrained and random-init protein sequence encoders at 35--150M parameters, we find that the best protein-JEPA design is not all-position latent prediction but a variant: predicting latent targets only at masked positions, and retaining the MLM cross-entropy. We call this recipe masked-position MLM+JEPA. On a 16-task downstream suite (15 frozen linear probes plus SCOPe-40 zero-shot fold retrieval), under matched wall-clock budgets, this recipe wins more tasks than it loses against MLM-only continuation: 10 wins / 3 losses / 3 ties (hereafter W/L/T) on pretrained ESM2-35M, 11/2/3 on ESM2-150M while results in pretraining from scratch are mixed (6/8/2). Gains are seen for multiple models on 11 of 16 tasks, including stability, \b{eta}\beta \b{eta}-lactamase fitness, variant effect, intrinsic disorder, remote homology, enzyme classification, and SCOPe-40 fold retrieval. Tasks with more losses than wins are Fluorescence (TAPE) and Peptide-HLA Binding. All-position MLM+JEPA matches MLM-only overall but does not reproduce the masked-position gains. JEPA-only (no MLM) collapses in nearly every experiment. We conclude that JEPA, when combined with MLM, is competitive and can outperform pure MLM in pretraining and continued training, even under matched wall-clock budgets.

cs.LG↗

arXiv2026

ProtSent: Protein Sentence Transformers

Protein language models (pLMs) produce per-residue representations that capture evolutionary and structural information, yet their mean-pooled sequence embeddings are not explicitly trained to reflect functional, evolutionary or structural similarity between proteins. We present Protein Sentence Transformers (ProtSent), a contrastive fine-tuning framework for adapting PLMs into general-purpose embedding models. ProtSent trains with MultipleNegativesRankingLoss across five protein-pair datasets: Pfam families, structurally derived hard negatives, AlphaFold DB structural pairs, and StringDB protein--protein interactions, and Deep Mutational Scanning data. We evaluate on 23~downstream tasks using frozen embeddings with a k-nearest-neighbor probe to measure embedding neighborhood quality. On ESM-2 150M, ProtSent improves 15 of 23 tasks, with gains of +105% on remote homology detection, +17% on variant effect prediction, and +19.9% Recall@1 on SCOPe-40 structural retrieval. The 35M variant improves 16 of 23 tasks with +40.5% on remote homology and +15.5% Recall@1 on SCOPe-40. Contrastive fine-tuning restructures the embedding space to better capture protein function and structure, without any task-specific supervision. We release the models, public data, and training recipe and code.

cs.LG↗

arXiv2025

InterFeat: A Pipeline for Finding Interesting Scientific Features

Finding interesting phenomena is the core of scientific discovery, but it is a manual, ill-defined concept. We present an integrative pipeline for automating the discovery of interesting simple hypotheses (feature-target relations with effect direction and a potential underlying mechanism) in structured biomedical data. The pipeline combines machine learning, knowledge graphs, literature search and Large Language Models. We formalize "interestingness" as a combination of novelty, utility and plausibility. On 8 major diseases from the UK Biobank, our pipeline consistently recovers risk factors years before their appearance in the literature. 40--53% of our top candidates were validated as interesting, compared to 0--7% for a SHAP-based baseline. Overall, 28% of 109 candidates were interesting to medical experts. The pipeline addresses the challenge of operationalizing "interestingness" scalably and for any target. We release data and code: https://github.com/LinialLab/InterFeat

q-bio.QM↗

arXiv2023

Whats next? Forecasting scientific research trends

Scientific research trends and interests evolve over time. The ability to identify and forecast these trends is vital for educational institutions, practitioners, investors, and funding organizations. In this study, we predict future trends in scientific publications using heterogeneous sources, including historical publication time series from PubMed, research and review articles, pre-trained language models, and patents. We demonstrate that scientific topic popularity levels and changes (trends) can be predicted five years in advance across 40 years and 125 diverse topics, including life-science concepts, biomedical, anatomy, and other science, technology, and engineering topics. Preceding publications and future patents are leading indicators for emerging scientific topics. We find the ratio of reviews to original research articles informative for identifying increasing or declining topics, with declining topics having an excess of reviews. We find that language models provide improved insights and predictions into temporal dynamics. In temporal validation, our models substantially outperform the historical baseline. Our findings suggest that similar dynamics apply across other scientific and engineering research topics. We present SciTrends, a user-friendly webtool for predicting scientific topics trends: https://hadasakaufman.shinyapps.io/SciTrend

cs.DL↗

arXiv2021

Open Problems in Human Trait Genetics

Genetic studies of human traits have revolutionized our understanding of the variation between individuals, and opened the door for numerous breakthroughs in biology, medicine and other scientific fields. And yet, the ultimate promise of this area of research is still not fully realized. In this review, we highlight the major open problems that need to be solved to improve our understanding of the genetic variation underlying human traits, and by discussing these challenges provide a primer to the field. Our focus is on concrete analytical problems, both conceptual and technical in nature. We cover general issues in genetic studies such as population structure, epistasis and gene-environment interactions, data-related issues such as ethnic diversity and rare genetic variants, and specific challenges related to heritability estimates, genetic association studies and polygenic risk scores. We emphasize the interconnectedness of these open problems and suggest promising avenues to address them.

q-bio.PE↗