Simon Axelrod (@AxelrodSimon) profile picture

Computational chemistry and machine learning @DEShawResearch | PhD at @HarvardCCB and @mit_dmse 2023 | @uoftphysics 2017 | @queensu 2016

Manhattan, NY
Joined March 2021
Photo by Simon Axelrod (@AxelrodSimon)
Love it. Clinical trial failures are still the biggest problem in drug discovery. Any progress here would be huge
We used AI to predict the failure of a Phase 3 trial before the results were announced. Today, we're publishing 10 more predictions for the future. Thread 🧵
6
385
Come say hi to the DESRES team! @DEShawResearch
At #NeurIPS23 this week with @DEShawResearch ! Stop by booth 421 if you wanna say hi!
1
23
3,304
Happy for the opportunity to present my latest work done in @RGBLabMIT at #ACSFall2023.
3
22
1,792
Simon Axelrod retweeted
Rafael Gomez-Bombarelli was one of twenty-six members of the MIT engineering faculty and researchers who received awards in recognition of their scholarship, service, and overall excellence in the past calendar quarter.
1
21
2,656
Simon Axelrod retweeted
How to optimize the per-site activity of oxide catalysts? Check out our work, "Atom-by-Atom Design of Metal Oxide Catalysts for the Oxygen Evolution Reaction with Machine Learning," led by our new PhD graduate, Jackie, and in collaboration with @EELabMIT! arxiv.org/abs/2305.19930
3
21
2,693
Jackie’s defense brought this amazing group of young people and big bright smiles together with @RGBLabMIT @mit_dmse @MITMechE @RLEatMIT
1
3
29
3,215
Use a VAE to predict atomic coordinates from coarse-grained coordinates (backmapping) for any protein system. @SoojungYang2 @RGBLabMIT arxiv.org/abs/2303.01569
29
125
11,831
Chemically Transferable Generative Backmapping of Coarse-Grained Proteins. (arXiv:2303.01569v1 [cs.LG]) #machinelearning #compchem cutt.ly/P8Pkyvc
5
12
1,624
Simon Axelrod retweeted
This weekend some of our group members braved the cold and enjoyed a winter hike at the Fells!❄️
3
1
27
3,318
Twitterverse - what does everyone think of this? Where does the gap between ML model performance and real world clinical trial improvements come from?
Replying to @wpwalters
Do you see a role for AI in reducing clinical trial failures? Models routinely get great scores on ClinTox, but afaik the clinical trial failure rate is still 90%. Are these models overfit and useless? Or are they great but ignored? Or something else?
4
3
7
4,985
Thermal half-lives of azobenzene derivatives: virtual screening based on intersystem crossing using a machine learning potential. NEW #ASAP by @AxelrodSimon, Eugene Shakhnovich, and @RGBLabMIT Read it here: go.acs.org/3NW
2
7
2,360
We use differentiable simulations to train machine learned interatomic potentials from radial distribution functions, opening a new avenue to derive force fields from experimental data. @RGBLabMIT @JDietschreit aippub.org/3Y46eRg
6
19
2,169
#compchem Good read: Learning pair potentials using differentiable simulations doi.org/10.1063/5.0126475
9
49
5,349
Simon Axelrod retweeted
Just out in ACS Central Science! We made an automated workflow for predicting the thermal half-lives of azobenzene derivatives. TL;DR isomerization involves intersystem crossing, multi-reference effects are critical, and ML gives fast and accurate results! pubs.acs.org/doi/full/10.102…
1
7
22
3,213
Simon Axelrod retweeted
Our recent work "Design principles for transition metal nitride stability and ammonia generation in acid" is featured #OnTheCover of @Joule_CP. Check out the paper using this free link (valid before 2/7/2023): authors.elsevier.com/a/1gHD0… @EELabMIT @mit_romangroup #Energy #ChemTwitter
4
6
59
7,400