Machine Learning | QTCOVI – Theoretical and Computational Chemistry

Machine Learning & Neural Network Potentials

Mon, 01 Jan 2024 00:00:00 +0000

Overview

Machine learning is transforming computational chemistry, and we are harnessing it in two complementary ways:

IQA-Informed Neural Network Potentials

Classical machine-learned interatomic potentials (NNPs) are trained on total energies and forces, but lack chemical interpretability. We develop NNPs informed by IQA energy components (self-energies and interaction energies), resulting in potentials that:

Decompose into physically meaningful atomic and pairwise contributions.
Transfer more reliably to out-of-distribution chemical environments.
Naturally encode the correct physics of bonding interactions.

Topological Descriptors as ML Features

QTAIM atomic properties and IQA energy components serve as physically motivated features for machine learning models targeting molecular properties, reaction barriers, and drug–target binding affinities.

Deep Learning for Electron Density

We explore the use of deep learning models to predict electron densities directly, enabling rapid computation of topological properties for large molecular datasets.

Codes & Tools

Our ML work builds on open-source frameworks (PyTorch, JAX) and is integrated with our in-house topological analysis codes.

MM2SF

Mon, 01 Jan 2024 00:00:00 +0000

About

MM2SF automatically generates optimised Atom-Centred Symmetry Functions (ACSFs) for use as descriptors in neural network interatomic potentials. Given a molecular dynamics trajectory or normal-mode sampling, it applies a Gaussian Mixture Model (GMM) to decompose the chemical space into well-defined clusters, then selects symmetry function parameters that accurately describe each region.

Supported symmetry function types:

Two-body (radial) — G^rad
Three-body (angular) — G^ang (modified functional form)

Installation

pip install git+https://github.com/m-gallegos/MM2SF.git

Or from a downloaded zip:

pip install MM2SF-main.zip

NNAIMQ

Mon, 01 Jan 2024 00:00:00 +0000

About

NNAIMQ predicts QTAIM (Bader) partial charges for C, H, O and N atoms in neutral, singlet-spin gas-phase organic and biological molecules. It comprises four Artificial Neural Networks (one per element) fitted to high-quality quantum chemical data.

Key features:

High-accuracy QTAIM charges without running a full topological analysis.
Supports standard .xyz geometry files as input.
Compatible with x86-64 and ARM (Apple M1) processors.

Requirements

Python ≥ 3.7.3
keras, matplotlib, numpy, pandas, seaborn, tensorflow

Usage

cd code/
python nnaimq.py input

where input is a plain-text file listing the .xyz geometry files to process.

SchNet4AIM

Mon, 01 Jan 2024 00:00:00 +0000

About

SchNet4AIM is a code designed to train SchNet deep-learning models on atomic (1-body) and pairwise (2-body) properties formulated within the Quantum Theory of Atoms in Molecules (QTAIM). It is built as a targeted modification of SchNetPack, retaining only the components relevant for 1p/2p property training.

Key features:

Train on atomic (charges, energies, volumes) or pairwise (delocalization indices, IQA interaction energies) QTAIM properties.
Supports JSON and ASE-SQLite database formats.
Runs on CPU and GPU (GPU recommended for speed).

Installation

git clone https://github.com/QTCOVI/SchNet4AIM.git
cd SchNet4AIM
pip install -r requirements.txt

NNAIMGUI

Sun, 01 Jan 2023 00:00:00 +0000

About

NNAIMGUI (M. Gallegos, University of Oviedo, 2023) is a code for the prediction and visualisation of atomic properties using feed-forward neural network (FFNN) models. It ships with the built-in NNAIMQ model for predicting QTAIM charges of gas-phase neutral singlet molecules containing C, H, O and N atoms, and supports user-supplied custom models for any atomic property of interest.

Key features:

Graphical user interface for non-expert users, plus command-line mode.
Built-in charge equilibration to enforce molecular electroneutrality (13 algorithms included).
Supports user-defined FFNN models for any atomic property.
Compatible with Linux and Windows.

Installation

pip install git+https://github.com/m-gallegos/NNAIMGUI.git

Citation

M. Gallegos et al., J. Chem. Inf. Model. (2023). https://doi.org/10.1021/acs.jcim.3c00597

Machine Learning | QTCOVI – Theoretical and Computational Chemistry

Machine Learning & Neural Network Potentials

Overview

IQA-Informed Neural Network Potentials

Topological Descriptors as ML Features

Deep Learning for Electron Density

Codes & Tools

MM2SF

About

Installation

Links

NNAIMQ

About

Requirements

Usage

Links

SchNet4AIM

About

Installation

Links

NNAIMGUI

About

Installation

Citation

Links