2024-02-21 FEB

DNA language models; gemini and/or openAI APIs?

Journal Club

Proc Natl Acad Sci U S A. 2023 Oct 31; 120(44): e2311219120. PMID: 37883436  [pdf]
DNA language models are powerful predictors of genome-wide variant effects

Gonzalo Benegas, a Sanjit Singh Batra, b and Yun S. Song b , c , d , 1
aGraduate Group in Computational Biology, University of California, Berkeley, CA 94720
bComputer Science Division, University of California, Berkeley, CA 94720
cDepartment of Statistics, University of California, Berkeley, CA 94720
dCenter for Computational Biology, University of California, Berkeley, CA 94720

Summary (Gemini)
Building language models for DNA is an emerging field with promising potential. These models, similar to how they work with natural language, can learn the patterns and relationships within DNA sequences and use this knowledge to perform various tasks, such as:

Several challenges exist in building DNA language models, including the vast amount of data required for training and the unique characteristics of DNA sequences compared to natural language. However, ongoing research is addressing these challenges, and the development of DNA language models is a rapidly evolving field.

Notes

Jeya's conversation with Gemini: https://gemini.google.com/share/55079d730ee6 

Hugging faces GPN models

https://github.com/songlab-cal/gpn/tree/main/analysis/human
https://github.com/songlab-cal/gpn/tree/main/analysis/arabidopsis 

Hachaton

Project updates

...

Gemini and/or openAI APIs?

...

epiVerse governance - that of an agent-based operating system.

Serverless Web Agents, mediated by token. 

Managing token is the critical question - https://ai.google.dev/pricing 

Either way generative AI OS is proposed as web agents

...