Making deep‑learning explanations for DNA more reliable
Official title: Reliable post hoc interpretations of deep learning in genomics
Funded at Cold Spring Harbor Laboratory · NIH-11333758 · Data updated 2026-11-12
This project improves computer methods that explain which parts of DNA control gene activity so researchers — and eventually patients — can better understand genetic effects.
Quick facts
| Grant type | R01 grant |
|---|---|
| Study type | NIH-funded research |
| Funding institution | Cold Spring Harbor Laboratory NIH-funded |
| Lab location | 1 site (Cold Spring Harbor, United States) |
| Project start | August 1, 2023 |
| Project end | April 30, 2027 |
| Funding year | 2026 |
| Award (this year) | $384,000 |
| NIH project number | 5R01GM149921-04 |
| Project ID | NIH-11333758 on NIH RePORTER |
Who could benefit from this research
✓Could help
This project does not appear to recruit patients directly; it mainly uses genomic datasets and collaborations with research labs rather than clinical enrollment.
✕Less likely to help
Patients seeking immediate treatment or clinical benefit are unlikely to gain direct, near‑term benefits from this computational methods project.
What this research studies
This work uses deep neural networks trained on genomic data to map which DNA letters matter for gene regulation and cell identity. The team will build methods to pick the best models, reduce noise in nucleotide‑level importance maps, and create benchmarks and best practices for interpreting results. They will test these tools across datasets such as ATAC‑seq and known transcription factor motif collections to ensure consistent, biologically meaningful outputs. The goal is clearer, more trustworthy maps of regulatory DNA that researchers can use to link genetic variants to function.
Why it matters
Potential benefit: If successful, this could help researchers pinpoint genetic changes that alter gene regulation and speed discovery of disease‑relevant variants and targets for therapy.
How similar studies have gone: Related deep‑learning approaches have improved prediction of regulatory DNA activity, but making their nucleotide‑level explanations reliable is still a largely open challenge.
Where this research is happening
-
Cold Spring Harbor Laboratory
Cold Spring Harbor, United States Active
Researchers
- Principal investigator: Koo, Peter K — Cold Spring Harbor Laboratory
- Study coordinator: Koo, Peter K
Related research
Other NIH-funded projects on the same topics.
-
Integrative histo-genomic analysis to guide cancer risk and targeted therapies with interpretable AI
Researchers are developing an advanced AI-powered approach to map genetic and cellular patterns across breast and cervical cancers to help tailor treatments.
-
Predicting how antimalarial drug resistance spreads with AI
A powerful AI model will be built to forecast how resistance to antimalarial drugs might spread in Africa and beyond, helping clinicians and public health teams stay ahead of outbreaks.
-
Self-supervised deep learning for biological sequences
A deep learning project uses self-supervised methods to better understand biological sequences and related data, with the goal of advancing biology and medicine.
-
Mobile phone imaging and AI to screen oral cancers in low-resource settings
A low-cost mobile phone device with artificial intelligence is being developed to help frontline health workers screen for oral cancer early in low-resource areas.
-
Linking Genetic Variants to Real-Life Health Effects with Smart AI
A new AI-powered approach will predict how genetic differences affect health across multiple levels to help doctors understand diseases and tailor care.
-
Integrative AI tools to map protein sequences to structures and functions
A scientist team will develop advanced AI methods to better connect protein sequences, their 3D shapes, and how they work, using large biological data from sequencing and structures.
About this research
- This is an active NIH-funded research project — typically early-stage science, not a clinical trial accepting patient enrollment.
- Some NIH-funded labs run parallel clinical studies or seek volunteers for related work. To check, contact the principal investigator or institution listed above.
- For full project details, budget, and progress reports, visit the official NIH RePORTER page.