Opinion: AI may never fully decode human genome regulation
The piece argues that, despite decades of sequencing, the human genome remains a complex regulatory puzzle that may be beyond the reach of current artificial‑intelligence approaches. While the Human Genome Project (1990‑2003) mapped the roughly 3 billion DNA bases, only about 2 percent codes for proteins, leaving the vast non‑coding regions to control gene activity. Researchers such as Adelman…
Key points
- Only ~2% of the human genome codes for proteins, the rest regulates gene activity
- Genomic foundation models such as Evo 2, Genos and AlphaGenome predict DNA effects but remain black boxes
- Gene regulation involves transcription factors, enhancers, chromatin loops and epigenetic marks in complex 3D structures
Genomic foundation models like Evo 2, Genos and Google DeepMind’s AlphaGenome are being used to predict DNA‑sequence effects, but the article cautions that these black‑box systems cannot capture the intricate, three‑dimensional dynamics of the genome. The author suggests that outsourcing analysis to AI may be useful for specific predictions, yet a full mechanistic understanding will likely require more than correlation‑driven models.
Overall, the author views the genome as a dynamic, physically folded system rather than a static code, implying that AI’s current input‑output paradigm may be insufficient for truly decoding human biology.
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 →- OpenAI fails to prove Navier-Stokes, mathematicians claim · 1 src
- You could have invented transformers from memory · 1 src
- Anthropic’s Claude Mythos 5 redesigns quantum circuits better than prior research · 2 src
- Danaher and NVIDIA partner to integrate AI into scientific instruments · 3 src
- LitGene transformer integrates text and ontology to improve gene representations · 1 src
Comments
via GitHub Discussions