Derek van Tilborg
@DerekvTilborg
Followers
686
Following
895
Statuses
97
PhD candidate in molecular fortune telling @TUeindhoven π³π± - @molecularML. Follow me on BlueSky: https://t.co/BPTUmQWs31
The Netherlands
Joined September 2018
Happy to finally share our latest paper on active deep learning in low-data drug discovery in @NatComputSci! We tried to answer some open questions to get ready for prospective application π§΅π #AI #chemistry #drugdiscovery
@fra_grisoni @molecularML
3
10
72
I have moved to bluer pastures. Follow me at BlueSky at
0
0
0
I want to delete my twitter account because of the Muskonian nature of this platform. What is the best alternative medium to stay in touch with the scientific community?
1
0
0
RT @agardinon: π¨Excited to share our new preprint: "Automated navigation of condensate phase behavior with active machine learning" https:β¦
0
8
0
A must read if you're doing molecular machine learning
0
1
6
A good read on the perceived success of molecular machine learning. Don't forget your statistical tests guys!
0
2
8
RT @fra_grisoni: π’ JOB ALERT π’ My team @molecularML is searching for a #postdoc with a background in medicinal chemistry! π©βπ¬π¨βπ¬π§ͺ Do youβ¦
0
47
0
We end the paper with some guidelines that came out of our systematic analysis. 1. Active learning will find you more hits, but you need to do several rounds of it 2. The acquisition method is by far the most important 3. ECFPs > GNNs 4. Biased start data isn't too problematic
0
1
6
De novo drug design was getting stale until @Rza_ozcelik came along with state space models to spice it up. Even though I'm totally biased, I think its really well done work :)
π₯³π₯³I'm thrilled that our work that introduces the state space models (SSMs) into de novo design is now published at @NatureComms π π w/ Sarah de Ruiter, Emanuele Criscuolo, and @fra_grisoni @molecularML
0
0
3
Playing with #AI is nice, but applying it is the real deal! Within 3 cycles, we optimized nanoparticle uptake in cancer cells by almost 15 fold. Cool for drug delivery, but mostly a lot of funπ§ͺ @AnaOrPe_ @rvdmnanomeds @fra_grisoni @Fluo_renzo paper:
Thrilled of our collaboration with the team of @Fluo_renzo, spearheaded by a great duo @AnaOrPe_ & @DerekvTilborg. π We show the integration of active #machinelearning, microfluidics, and imaging for nanoparticle design. Just out on @digital_rsc πͺπ»
2
2
20
Our (mini) review on deep learning for low-data drug discovery is now out! There are honestly few things I enjoy as much as writing a review. It's so satisfying Especially together with @hlnbrnkmnn @emanuelecriscu9 Luke Rossen @Rza_ozcelik @fra_grisoni
Check out our new review on "Deep learning for low-data #drugdiscovery"! We discuss ML strategies to deal with our most pressing problem; β¨ low-data β¨ @hlnbrnkmnn @emanuelecriscu9 Luke Rossen @Rza_ozcelik @fra_grisoni @molecularML
#AI #chemistry
1
19
53
@Ryan__Rhys @hlnbrnkmnn @emanuelecriscu9 @Rza_ozcelik @fra_grisoni @molecularML Thanks! Glad you like it. We were bound to a bit of a word limit with this one, but proper uncertainty estimation is extremely insightful in low-data settings. (I'm very much into it personally)
1
0
1
@WowThatBurns @hlnbrnkmnn @emanuelecriscu9 @Rza_ozcelik @fra_grisoni @molecularML I would say that FS-mol is the only purpose-built truly low-data benchmark there is as far as I know
0
0
2
@fra_grisoni had the idea of writing this review in 1 day with the group. We all came prepared and crunched out a messy draft in one work day. The rest definitely took more than a day (a lot longer), but it was a nice kickstart and teambuilding event :)
0
0
2
We cover three main topics: 1. Data augmentation, where we talk about how to augment molecular data.
1
0
1