Tweets

2023-07-06T14:13:23+00:00 | 🔗

I didn't realize the original plot's x-axis is 0-0.8 not 0-1, so the distribution from non-DNN approaches is actually much closer EXCEPT for that last bit from 0.67-0.8 which is also an important bit. https://t.co/zqKyzZeAMA

2023-07-06T13:47:44+00:00 | 🔗

It gets stuck at a weird maxima of ~0.67 similarity (blue), any ideas for breaking out? PS, this is the original papers results (yellow) PPS, contrasting with original approach, this starts from a random population not the patent population https://t.co/RNsoQoQILy

2023-07-06T13:47:43+00:00 | 🔗

Inspired by the Digital Discovery paper from @RGBLabMIT, I wondered how close can you get to Erlotinib using Tanimoto similarity reward like in the paper. So we started with a 2000 compound sample of ZINC22 and used CrEM, C-C slicing, and GB-GA and ran it for 25 generations. https://t.co/ntkF3iUxdj

2023-07-05T15:15:18+00:00 | 🔗

RT @datamol_io: Today, we are excited to introduce Graphium into the https://t.co/V1RU6QWReZ ecosystem! Graphium is a powerful and flexibl…

2023-07-04T05:18:04+00:00 | 🔗

Don't know why the quality of that first workflow img is, here it is https://t.co/RFIgOkRJSG

2023-07-04T05:15:52+00:00 | 🔗

Using WHALES (for scaffold-hopping) + SynSpace (for reaction-based modifications) to generate a torrent of hypotheses, which are mostly synthetically reasonable (guess what the starting compound was) https://t.co/ZR0d6v7JOn

2023-07-01T18:35:03+00:00 | 🔗

Well I’m off Twitter now, I don’t even think I can see my own tweets even those exceed the rate limit.

2023-06-29T17:47:56+00:00 | 🔗

@AnthonyLeeZhang Wait till they start offering to protect your cars for a recurring payment, but then you have multiple people wanting protection money, and then they compete and then organize, and voila

2023-06-28T14:52:17+00:00 | 🔗

@andrewwhite01 @payel791 I think here: https://t.co/GhXsPXLvaQ

2023-06-24T17:47:31+00:00 | 🔗

A weird way to get out of Ukraine.