How AI Learned to Read the Genomic “Dialects” of 300,000 People for Precise Expression Prediction

This article reviews a study that overcomes the limitation of reference‑genome‑only models by pre‑training a genomic language model on 300,000 European individuals’ variants, creating UKBioBERT and the two‑stage UKBioFormer, which together deliver markedly better gene‑function representations and personalized expression predictions across cell lines and populations.

UKBioBERTUKBioFormerfunctional genomics

0 likes · 7 min read

How AI Learned to Read the Genomic “Dialects” of 300,000 People for Precise Expression Prediction

HyperAI Super Neural

Nov 10, 2025 · Artificial Intelligence

Columbia & Stanford Launch Squidiff: Diffusion Model for Transcriptome Simulation

Squidiff, a conditional diffusion framework co‑developed by Columbia and Stanford, predicts transcriptional responses across cell differentiation, gene and drug perturbations, and radiation exposure, outperforming prior models and enabling more precise and spatially aware biomedical research.

AI for BiologyPerturbation PredictionSingle-Cell Transcriptomics

0 likes · 16 min read

Columbia & Stanford Launch Squidiff: Diffusion Model for Transcriptome Simulation

Data Party THU

Sep 21, 2025 · Artificial Intelligence

How Delphi-2M Uses Generative Transformers to Predict Over 1,000 Diseases

A new AI system called Delphi-2M, built on an enhanced generative‑transformer architecture and trained with UK Biobank data, can forecast the risk of more than a thousand diseases up to twenty years in advance, achieving an average AUC of 0.67 in external validation while offering explainable, extensible predictions for personalized health.

AIDelphi-2MDisease Prediction

0 likes · 5 min read

How Delphi-2M Uses Generative Transformers to Predict Over 1,000 Diseases