Highlights of the Year 2024! Last Weekly Update?

Another year flew by like a comet. And what a year it was! This year was packed full of updates from my group and, of course, MLatom, which is at the center of our research. MLatom provides us with a tool to develop better AI-enhanced methods for computational chemistry while at the same time serving as a platform to enable such simulations.

Every Wednesday in 2024, we released a weekly video to share with you our updates and tutorials on how to use MLatom. We committed to this and kept the streak for a year, even on holidays. Hey, today is Christmas, and we are releasing a new video again! I have not yet made my decision whether I want to keep doing it next year. I might redirect my energy in a very different direction, where videos might play a part but in a very different format. Please let me know your thoughts by leaving a comment or drop me a message.

Back to the year recap, it was the most productive year in our group’s research. We developed a host of new powerful methods and started to more intensively apply them. You can check out the full list of our 20+ papers and preprints in 2024, while here I just recap 5 6 of them:

1. UAIQM – the biggest breakthrough of our recent years in terms of creating a powerful platform for AI-enhanced computational chemistry, which you can use out of the box for your simulations to get results faster and more accurately than with many common DFT approaches. UAIQM is becoming increasingly popular but the paper remains a preprint, which often happens for breakthrough works.
2. DENS24 – the paper title says it all: “The best DFT functional is the ensemble of functionals”. It was the most fun work which has been published in a cool journal Advanced Science.
3. Charting electronic-state manifolds across molecules with multi-state learning and gap-driven dynamics via efficient and robust active learning. It is also still a preprint, but for me it holds a special place as it was the result of 8-year efforts to create a practically useful ML tool for nonadiabatic dynamics.
4. Active learning tool with physics-informed uncertainty quantification and automatic setups which led to a fun discovery of roaming in Diels–Alder reaction with fullerene C₆₀ (ok, that’s one extra paper).
5. MLatom 3 paper in one of the best journals in our field: JCTC. It was published at the beginning of the year and got quite outdated as we released more than 20 versions of MLatom since.

Talking about over 20 releases of MLatom (minor and major), if you have not upgraded it yet – please do so now:

pip install --upgrade mlatom

As you can see from a list of updates on the GitHub page of MLatom, it grew very fast. You can use it for various simulations mentioned above (UAIQM, DENS24, active learning for ground- and excited-state simulations) and also for much more such as accurate and fast IR and Raman spectra simulations, fine-tuning of the universal models, analysis of the calculations, and so on.

I should mention that this was also the year, when I started using MLatom as a fully-fledged program for teaching AI and computational chemistry, in the universities and workshops. This led to the development of an online course.

I am very happy to take this opportunity to express my big thanks to everyone contributing to all this amazing progress in one year: MLatom developers and users, and XACS team bringing our developments to the online platform.