Conversation

"machine learning" 🙄 you still have to do the testing & gather the data. what did the machine learn on? models? I'm just not seeing why you need AI as opposed to a searchable database for chemicals with properties you're looking for.
6
1
Because it’s not just “apply chemical X on target Y for effect Z”. It’s developing a new compound & hoping it has a particular effect in an incredibly complex system. Think of it like an optimization problem, not a password to be cracked for which a rainbow table already exists.
19
This paper is so cool!!!! More legit than Halicin, but not what I would describe as a full active/in-silico system... ... Abaucin is a previously investigated compound that was flagged for inhibition potential by the active learning system in this paper...
1
8
Show replies
The problem with antibiotic drug discovery is that it is unprofitable. There have been many, many exciting molecules developed at the laboratory/pre-clinical stage over the past 30 years. They are not brought to market because it is unprofitable to do so.
2
6
Unknown. As with all therapeutic development, the proof is in human clinical trials. Many things work in silico, in vitro, and in vivo with animals, that ultimately fail in humans. The phrasing they used in the abstract, "promising lead," is correct. Hope it works!
13
the ml is sound, basically just finding factors that suggest effectiveness against the bacteria and then iterating based on those factors, but idk the bio side, how feasible it is to produce the new molecule, if it’s actually effective in animals etc
1