2
$\begingroup$

We've had a few questions recently about deploying large language models (LLMs) for text interrogation of biological publications. Essentially these questions are "where do I start"? In other words, PI says "I've heard of LLMs, okay student project. No idea what it's about but here's the theme" and the student posts here. In my personal view this ain't ideal academic line management.

Anyway, this isn't bioinformatics, so these posts are off-topic. This is because the primary data are text files, probably publication pdfs and whilst the theme of the data is biology, the data ain't a result of biological wet-lab experimentation. My personal view is these questions should be directed to Data Science Stackexchange and thats where I've transferred them previously.


Just for the record, I do know about LLMs. I've trained in them - not used in production. My job's analytics and LLMs are intriguing. Thus, IMHO they're great, potential huge, but they ain't bioinformatics.

$\endgroup$
2
  • 1
    $\begingroup$ I don't think you can completely discount LLMs as "not bioinformatics", but in of themselves as a text mining tool it is not appropriate. For example, this paper describes using LLM tools to query spatial transcriptomic data and gain insight about the biological nature of that data; that sits squarely in the realm of bioinformatics. $\endgroup$
    – gringer Mod
    Commented Jul 15 at 22:32
  • $\begingroup$ Thats a good point @gringer, however Chao Hui Huang isn't published. Whilst NLP can be used for oligo-peptides supervised learning, LLMs are very different and very focused towards human languages via their GAN construction. However, if I'm wrong, I'm wrong. $\endgroup$
    – M__ Mod
    Commented Jul 16 at 2:35

0

You must log in to answer this question.

Browse other questions tagged .