this post was submitted on 22 Aug 2023
536 points (98.4% liked)
Asklemmy
43945 readers
792 users here now
A loosely moderated place to ask open-ended questions
Search asklemmy ๐
If your post meets the following criteria, it's welcome here!
- Open-ended question
- Not offensive: at this point, we do not have the bandwidth to moderate overtly political discussions. Assume best intent and be excellent to each other.
- Not regarding using or support for Lemmy: context, see the list of support communities and tools for finding communities below
- Not ad nauseam inducing: please make sure it is a question that would be new to most members
- An actual topic of discussion
Looking for support?
Looking for a community?
- Lemmyverse: community search
- sub.rehab: maps old subreddits to fediverse options, marks official as such
- !lemmy411@lemmy.ca: a community for finding communities
~Icon~ ~by~ ~@Double_A@discuss.tchncs.de~
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I build and optimize molecules using computational chemistry software for fun.
Tell me more. I'm no rocket surgeon, so use comprehensively monosyllabic word thingys.
I've read that some do it to help the scientific community figure out stuff. What though. Idk.
Is AI doing this too? Is it faster, better, more creative at optimizing?
Aren't molecules already optimized, so what are you actually doing? Folding them? What does that even mean?
Help me understand.
Molecules that have been made are. Molecules that I investigate havent been made yet. They're drawn in jmol by hand and need to have their structures optimized otherwise you can't accurately calculate those molecules' properties. Eg. Raman spectra, strength of interactions between them and other molecules or ions, what the orbitals look like i.e how the molecule is actually held together or in the case of plane wave calculations and with the assistance of other software, you can design and build new materials with interesting properties. Eg. Stronger materials, materials that exhibit linearity in transistors (nanotubes are an example of this but so far various issues with their manufacture has not lead to strong evidence of this in experiments for a number of reasons)
New materials, catalysts to speed up reactions, new drugs, this sort of work has a lot of useful applications but its also just interesting to see what can be built just for the hell of it.
That's pretty cool. A unique type of puzzle to solve
A puzzle is a pretty good analogy for designing drug molecules. Drugs interact with proteins and enzymes much like trying to snap a puzzle piece into a puzzle. i.e lock and key model. Currently the strategy drug companies use essentially amounts to synthesizing millions of potential drug candidates in small quantities and testing for target activity. This is... horribly inefficient and it is hoped that better modeling can help cut down on either the number of candidate drugs that need to be screened and/or refine the drugs' activity to be more targeted (reduce side effects)
Thank you for explaining it to a layperson. Your enjoyment of this hobby shines through and so you've enlightened me to some new in the world.
Has your hobby produced anything useful or sharable? Or is like all the beautiful pictures I've taken that remain on my phone?
I am working toward submitting some of it for publication in a peer reviewed scientific journal. The research in question involves new aromatic ring systems that mostly have applications as catalysts.
LAMMPS?
Nwchem, orca, gamess and occasionally I tutor students in the use of gaussian.
I used to mess with LAMMPS in college, and now that I have pursued software engineering, I wish I could get back into it and see how I can contribute.
Any resources you'd recommend to get back into this?
Into lammps or in general? Because I havent really messed with LAMMPS so I wouldnt be able to reccommend resources for that yet. But if you mean in general, orca and gaussian are the easiest to use. Nwchem and gamess are a bit more obtuse with their documentation.
Chemcompute autogenerates input files and can be used to run simple calculations on a supercomputer but is limited to 1 hour with 1 core unless you are affiliated with their parner universities. In the latter case you can use up to 32 cores for longer periods of time. Currently it has support for gamess, tinker, jupyterhub, psi4 and namd: https://chemcompute.org/
ORCA has a wiki for input file documentation and use: https://sites.google.com/site/orcainputlibrary/visualization-and-printing
Avogadro, jmol and chemcraft are common drawing programs that can be used to draw molecules and visualize various properties. (ORCA uses .out as the extension of the output file, gaussian uses cube files and nwchem can be told to output data as cube files but again, is much more obtuse)
Avogadro manual: https://avogadro.cc/docs/ Nwchem:https://nwchemgit.github.io/Home.html Gaussian: https://gaussian.com/man/ Gamess: https://www.msg.chem.iastate.edu/gamess/documentation.html