<i>K</i>_DEEP: Protein–Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks

TLDR

Accurately predicting protein–ligand binding affinities is an important problem in computational chemistry since it can substantially accelerate drug discovery for virtual screening and lead optimization. The study proposes a fast machine‑learning approach using 3D‑convolutional neural networks to predict binding affinities and compares it to other machine‑learning and scoring methods across diverse datasets. The approach employs 3D‑convolutional neural networks trained on diverse datasets and is implemented in KDEEP, available on PlayMolecule.org, where each prediction takes a fraction of a second. On the PDBbind v.2016 core test set, KDEEP achieves a Pearson correlation of 0.82 and RMSE of 1.27 pK, outperforming other methods, though performance varies by protein, and its speed and ease of use make it attractive for computational chemistry pipelines.

Abstract

Accurately predicting protein–ligand binding affinities is an important problem in computational chemistry since it can substantially accelerate drug discovery for virtual screening and lead optimization. We propose here a fast machine-learning approach for predicting binding affinities using state-of-the-art 3D-convolutional neural networks and compare this approach to other machine-learning and scoring methods using several diverse data sets. The results for the standard PDBbind (v.2016) core test-set are state-of-the-art with a Pearson's correlation coefficient of 0.82 and a RMSE of 1.27 in pK units between experimental and predicted affinity, but accuracy is still very sensitive to the specific protein used. KDEEP is made available via PlayMolecule.org for users to test easily their own protein–ligand complexes, with each prediction taking a fraction of a second. We believe that the speed, performance, and ease of use of KDEEP makes it already an attractive scoring function for modern computational chemistry pipelines.