Teaching an old dog new cells

2026-05-25 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Research Methodology & Innovation · Depth: Expert, quick

Summary

Two new methods have been developed to efficiently adapt bulk-trained sequence models for predicting gene regulation and variant effects at single-cell resolution. These advancements, detailed in an article titled "Teaching an old dog new cells" by Johannes Linder and David R. Kelley of Calico Life Sciences, LLC, published in Nat Methods on May 25, 2026, represent a significant step in applying existing computational models to more granular biological data. The core innovation lies in enabling models typically trained on aggregated data to perform high-resolution analysis, offering a more precise understanding of cellular processes. This capability is crucial for advancing genomic research by providing detailed insights into how genetic variations influence individual cell behavior and function, moving beyond population-level averages.

Key takeaway

For research scientists working with genomic data, these new methods mean you can now apply existing bulk-trained sequence models to achieve single-cell resolution for gene regulation and variant effect prediction. This significantly enhances the precision of your analyses, moving beyond population-level averages to individual cellular insights. Consider exploring these adaptation techniques to refine your current genomic modeling workflows and uncover more granular biological mechanisms.

Key insights

Bulk-trained sequence models can be efficiently adapted for single-cell gene regulation and variant effect prediction.

Principles

• Existing models can gain new resolution.

In practice

• Predict gene regulation at single-cell level.
• Analyze variant effects in individual cells.

Topics

• Sequence Models
• Gene Regulation
• Variant Effects
• Single-Cell Genomics
• Transfer Learning
• Computational Biology

Best for: AI Scientist, Research Scientist

Related on AIssential

• Toward informed batch correction for single-cell transcriptome integration — Batch effects in single-cell data hinder comparisons; future methods need informed, interpretable modeling.
• Decoding sequence determinants of gene expression in diverse cellular and disease states — Decima predicts cell-type-specific gene expression from DNA sequence, enabling precise regulatory mechanism and variant effect analysis.
• [R] Genomic Large Language Models — Genomic LLMs can identify functional similarities in DNA beyond sequence alignment, indicating learned regulatory patterns.
• The New DNA Problem: We Can Read It, But Can We Understand It? — AI is transforming complex genomic data from next-generation sequencing into understandable insights for non-experts.
• Probabilistic modelling of single-cell bisulfite sequencing data with MethylVI — MethylVI disentangles biological signals from technical noise in single-cell bisulfite sequencing data.
• Harnessing the power of single-cell large language models with parameter-efficient fine-tuning using scPEFT — scPEFT efficiently adapts single-cell LLMs to new tasks, reducing computational demands and improving performance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.