Unified imputation of missing data modalities and features in multi-omic data via shared representation learning www.biorxiv.org/content/10.64898/2026.02...

The idea is to learn a shared representation across modalities that supports imputation in a unified way, instead of treating “missing modalities” and “missing values” as two separate problems.

What do you do when you don’t have all modalities for all samples in multi-omic data, and the missingness is messy?
We developed a method for this.
MIMIR: Unified imputation of missing data modalities and features in multi-omic data via shared representation learning
www.biorxiv.org/content/10.6...

ML-Guided GWAS Reveals Genetic Architectures for MASLD for Overweight and Lean Individuals in the All of Us Cohort www.medrxiv.org/content/10.64898/2025.12...

Protein language models can predict epistasis (non-additivity of multiple point mutations) if predictions are nonlinearly transformed first www.biorxiv.org/content/10.1...

Protein Language Models Capture Structural and Functional Epistasis in a Zero-Shot Setting www.biorxiv.org/content/10.1101/2025.09....

We were wondering: Do protein language models learn epistatic signals during pretraining?
Turns out, yes!
Protein Language Models Capture Structural and Functional Epistasis in a Zero-Shot Setting
www.biorxiv.org/content/10.1...