🧬

Oh no argument here. Truly just pointing out that other molecules are technically supported. Excited to see how it develops in the future. Definitely a lot of potential!

👍

The really cool part is that it's currently the worst it's ever going to be. It will only get better from here!

This is incorrect. While proteins are the focus and the star of the show, Alphafold does accept DNA, RNA, and even small molecule ligands as inputs. It even outperforms SOTA alternatives on RNA structures according to some metrics. Not as effective on those, to be sure, but it is not 'protein only.'

Great thread. It seems like there's a lack of awareness of the sheer breadth of AI applications among the general public. Image generators and chatbots = AI = bad. But in reality it's a conversation that needs nuance, and condemning "AI" unilaterally may do far more harm than good.

Protein LMs, genomic LMs, and chemical LMs, to name only a few. Incredible potential to transform medicine (already are), and they're based on the same tech as any language model. And all are technically subsets of machine learning. So, not completely different, to be fair.