Researchers release Tangermeme toolkit for interpreting genomic deep learning models
A team of researchers has released Tangermeme, an open-source Python toolkit designed to help scientists interpret what genomic deep learning models have learned about cis-regulatory logic. Published in Nature Machine Learning, the software focuses on "everything-but-the-model" — providing optimized, modular tools for sequence manipulation, prediction, feature attribution, and sequence design…
Key points
- Tangermeme is an open-source toolkit for interpreting genomic deep learning models, released under MIT license
- Includes novel recursive seqlet caller that identifies variable-length regulatory motifs from model attributions
- Demonstrated on BPNet and Beluga models showing different learned cis-regulatory logic for MYC binding
Tangermeme introduces a novel recursive seqlet-calling algorithm that identifies variable-length sequence motifs driving model predictions, along with tools for annotating these motifs against databases like JASPAR2024 and analyzing their pairwise occurrences and spacing. The authors demonstrate the toolkit by comparing BPNet and Beluga models trained on MYC transcription factor binding, revealing that BPNet relies almost exclusively on MYC motifs while Beluga learns a broader regulatory vocabulary. The package also includes optimized one-hot encoding that processes the entire human chromosome 1 in under 2 seconds, and built-in batching for DeepLIFT/SHAP attribution that avoids GPU memory bottlenecks. Tangermeme is available under the MIT license via pip and GitHub, with documentation at tangermeme.readthedocs.io.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Coding Agents Can Solve the Task — But How Much Work Did They Waste? · 1 src
- Anthropic research suggests small AI models may self-replicate on IoT devices · 5 src
- Sakana AI's Multi-Layered Review catches 73.43% of core-claim errors · 1 src
- Anthropic says anti-distillation measures may have cut data harvested by 90‑99% · 2 src
- Is this idea from AI or humans? A detector claims it can tell · 1 src
Comments
via GitHub Discussions