We did a Biology x ML hackathon

[@ASandwichNinja]

I'm a chemist and, while I won't pretend to understand most of the content of machine learning from this video, I think that is a really interesting project. The idea of weighting the importance of certain fragments makes sense to me as that's kind of the human algorithm for looking at the mass spectra of molecules anyway. During the fragmentation/ionisation process there are certain structural motifs that are much more stable than others and so more likely to form (the one I always remember from my undergrad courses on mass spec is the tropylium ion being exceptionally stable as an aromatic cation). I wonder if programming in explicit examples of smaller fragment ions resulting from known mechanistic routes from common fragments could aid in the initial training of the model but I'm no expert in ML. In any case, really interesting video. One correction I would make is that the common name for the technique is mass spectrometry, not mass spectroscopy. The definitions of spectrometry vs spectroscopy get more opinion based the more you look into it in my experience, but a commonly held belief is that spectroscopy involves the measurements of interactions between materials and electromagnetic radiation whereas mass spectrometry directly measures properties (mass to charge ratio) of ion fragments. A bit pedantic but I've met a fair few people who really get hung up on this stuff.

[@thebiblioholic]

Very cool! And loved your submission. So much more impactful with hand-drawn slides than fancy graphics like every other powerpoint deck. I've been a judge for high school science projects for many years now, and I'm seeing some very sophisticated ML projects each year.

[@michaelsommers2356]

Don't worry about not being a biologist. Many of the first generation of molecular geneticists were actually physicists, including Francis Crick. Schoedinger, when he wasn't killing cats, wrote a book called _What is Life?,_ which inspired a generation or two of biologists, including Watson. There is a book by Judson called _The Eighth Day of Creation_ which give an historical account of the field, and which also talks about the contributions of many physicists.

[@cocoscacao6102]

Having fun (read stress) and learning something new is the main point of hackathons. Plus, you might meet new people.

[@mrelectro-istudy]

just saw your fourier series video it was really good thanks :)

[@universalsoup630]

That's really cool! It'd be great if you could suggest some other (maybe less selective) hackathons for people interested in giving something like this a go!

[@LookingGlassUniverse]

Awesome, I’m so glad you’re considering it! Most hackathon’s aren’t selective I think. I think this one was just because they gave us free credits to work with. The article in the description mentioned a lot of very cool bio ones. The coolest one is ending next week: Adaptiv Bio has a competition to design proteins which they will then make in the lab. Try it out if you can!

[@LarsBahner]

You are my Hero. I love how you dare to try things and and are not afraid of failing. After all - thats how you learn and get better. Doing the double slit and a hackathon like is something more thinkers and thinkers should dare to do.❤🎉

[@cianjones2728]

Great video, would love to see an in depth video about this topic

[@platypi_otbs]

That's cool. You're extremely brilliant.

[@sorin.n]

Amazing work!

[@LumTheAlien]

ChemBERTa is not a program I had heard of before, Thank you for the information.

[@loganbeavers9799]

That's pretty damn cool

[@jiteshvora86]

6:37 - How did you get 8 million parameters on the final layers? Isn't 400k x 10 fragments = 4 million?

[@LookingGlassUniverse]

2 numbers per fragment: the mz and intensity

[@jiteshvora86]

@@LookingGlassUniverse I see, that makes sense, thanks!

[@davidaugustyn9234]

Did you learn any chemistry for your physics degree or

[@mutabazimichael8404]

Fascinating

[@dennis4248]

Can I book lessons with you? You seem like a wonderful teacher

[@Cipherspace02]

im nobody but i always fear hackathons ( im from mechanical branch ) but hackathons are really cool

[@giovannicianci4961]

❤

[@pauldirac6243]

Have you totally given up your physics career? Where's the QM?

[@markxxx21]

This person constantly underestimates herself. Her strongest attribute is to break down the complicated for her KZbin audience. Her video on spin is amazing if you still aren't getting it, which apparently a lot of physicists don't.

[@linusbao8650]

As someone who just finished their thesis on Graph Neural Networks for molecular property prediction I was practically screaming at my screen the entire video begging you to use a GNN. Glad to see a GNN mentioned at the end of the video 🥲