Orbital Mixer: Using Atomic Orbital Features for Basis-Dependent Prediction of Molecular Wavefunctions

Abstract

Leveraging ab initio data at scale has enabled the development of machine learning models capable of extremely accurate and fast molecular property prediction. A central paradigm of many previous studies focuses on generating predictions for only a fixed set of properties. Recent lines of research instead aim to explicitly learn the electronic structure via molecular wavefunctions, from which other quantum chemical properties can be directly derived. While previous methods generate predictions as a function of only the atomic configuration, in this work we present an alternate approach that directly purposes basis-dependent information to predict molecular electronic structure. Our model, Orbital Mixer, is composed entirely of multi-layer perceptrons (MLPs) using MLP-Mixer layers within a simple, intuitive, and scalable architecture that achieves competitive Hamiltonian and molecular orbital energy and coefficient prediction accuracies compared to the state-of-the-art.

DiAMoNDBack: Diffusion-Denoising Autoregressive Model for Non-Deterministic Backmapping of Cα Protein Traces

“…Conditioning for each step of the residue-by-residue autoregressive backmapping is based on the Cα trace, the N = 14 most spatially proximate residues (i.e., those residues that have been brought into proximity by a secondary structural element, the tertiary fold, or quaternary complex, but which may be distantly separated in the backbone amino acid sequence), and a one-hot encoding of the residue type. Backmapping is performed by aligning each residue into a canonical alignment that permits us to directly generate the Cartesian coordinates of each backmapped atom in a rotationally and translationally invariant reference frame that avoids the need for costly data augmentations otherwise required to implicitly learn insensitivity to rigid atomic rotations and translations. , …”