Explainable AI identifies a new class of antibiotics, finishing news from NeurIPS 2023, delivering AI Assistants, dearraying update

Rebecca Sansale hackathon challenge

https://ncbiinsights.ncbi.nlm.nih.gov/event/ml-ai-solutions-for-biological-research-codeathon 

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE98965 

Register for Quarto (https://bioinformatics.ccr.cancer.gov/btep/classes/documenting-your-analysis-with-quarto) at 11:00-noon

promote https://quarto.org/docs/computations/ojs.html as the common ground between stats and data science

Jeya's selection 

of presentations of Quarto more focused on what we are focused:

https://jthomasmock.github.io/quarto-2hr-webinar

https://www.youtube.com/watch?v=yvi5uXQMvu4

Journal Club

Summary paper: https://www.nature.com/articles/d41586-023-03668-1 

NATURE RESEARCH BRIEFINGS
20 December 2023

‘Explainable’ AI identifies a new class of antibiotics

An artificial-intelligence graph neural network was trained on experimental data and used to identify chemical substructures that underlie selective antibiotic activity in more than 12 million compounds. This led to the discovery of a class of antibiotics with in vitro and in vivo activity against Gram-positive bacteria, including Staphylococcus aureus.

Full paper: https://www.nature.com/articles/s41586-023-06887-8 

______________________________________________________

... finish the last two weeks of

NeurIPS 2023 report by Jeya

______________________________________________________

Memory lane: preprints before preprints

Data points on (FAI)Reusability crisis:

Remember reproducibility, (https://www.nature.com/articles/d41586-021-01174-w) see 2023 tally of 10,000 manuscript retractions https://www.nature.com/articles/d41586-023-04035-w 

Hackathon

Delivering AI Assistants

recall last FAIR Friday, and the use of https://platform.openai.com/playground , can/should we poxy governance?

Show and tell
Please share your assistants :-)

The e-Utils use case:
https://www.ncbi.nlm.nih.gov/books/NBK25501 (manual) (quick try)
// e-Utils
// txt query --(eSearch)--> UIDs
// UIDs --(eSummary)--> record eSummary
// UIDs --(eFetch)--> full record
// UIDs in db A --(eLink)--> UIDs in db B
// UIDs --(ePost)--> temporary storage

DeArraying update

Creating standards, acting on them remotely [Aaron]

epiVerseUI

... if there is time to recap for Phillip, including Lee's observable angle:

The GPT hook could be a function, ideally a backtick function as Observable users are familiar with that notation for HTML and Markdown). So we could call an LLM like:

1

gpt`List the drugs which treat asthma` 

The function would handle the GPT calling loop, and respond with a basic chat UI cell that allows the user to inspect the function calling chain, respond to the messages, etc. The value of the cell would be any data the function gathered, immediately hooking the function calling results into the reactive Observable environment.