Chem. Commun. Feature Article: “AI in computational chemistry through the lens of a decade-long journey”

My review ‘AI in computational chemistry through the lens of a decade-long journey’ was published open access as an invited Feature Article in Chemical Communication. It gives a perspective on the progress of AI tools in computational chemistry through the lens of my decade-long contributions put in the wider context of the trends in this rapidly expanding field.

After reading the review, you will learn about:

• why you should use ML-improved QM methods whenever possible (e.g., AIQM1 instead of B3LYP for neutral, closed-shell CHNO-containing molecules)
• the power of Δ-learning providing a robust solution for integrating ML with QM and …
  • how it is related to transfer learning
  • how it can be generalized to learning from multiple levels of QM methods (hierarchical ML)
  • why you should use the term hierarchical ML and not Δ-learning when your baseline is also ML potential
• the zoo of machine learning potentials
• software for AI-enhanced computational chemistry and democratizing these calculations through the XACS cloud computing
• running faster ground– and excited-state dynamics, UV/vis and two-photon absorption spectra
• how AI enables fundamentally new ways of simulations, e.g., by directly learning MD trajectories
• ChatGPT and other elephants in the room
• and much more.

A little background story – I was very pleased to be invited to write a review for this respectable journal, as ten years ago I was trying to submit a paper there as a single author with just an M.Sc. degree. That paper got published in another journal though.

I thank everyone who shaped my research during these years!

Paper

Pavlo O. Dral*. AI in computational chemistry through the lens of a decade-long journey. Chem. Commun. 2024, 60, 3240–3258. DOI: 10.1039/D4CC00010B. (open access under the CC-BY license)

Feature Article in Chemical Communications