CAPLA: improved prediction of protein–ligand binding affinity by a deep learning approach based on a cross-attention mechanism

Abstract

Motivation Accurate and rapid prediction of protein-ligand binding affinity is a great challenge currently encountered in drug discovery. Recent advances have manifested a promising alternative in applying deep learning-based computational approaches for accurately quantifying binding affinity. The structure complementarity between protein-binding pocket and ligand has a great effect on the binding strength between a protein and a ligand, but most of existing deep learning approaches usually extracted the features of pocket and ligand by these two detached modules. Results In this work, a new deep learning approach based on the cross-attention mechanism named CAPLA was developed for improved prediction of protein-ligand binding affinity by learning features from sequence-level information of both protein and ligand. Specifically, CAPLA employs the cross-attention mechanism to capture the mutual effect of protein-binding pocket and ligand. We evaluated the performance of our proposed CAPLA on comprehensive benchmarking experiments on binding affinity prediction, demonstrating the superior performance of CAPLA over state-of-the-art baseline approaches. Moreover, we provided the interpretability for CAPLA to uncover critical functional residues that contribute most to the binding affinity through the analysis of the attention scores generated by the cross-attention mechanism. Consequently, these results indicate that CAPLA is an effective approach for binding affinity prediction and may contribute to useful help for further consequent applications. Availability The source code of the method along with trained models are freely available at https://github.com/lennylv/CAPLA. Supplementary information Supplementary data are available at Bioinformatics online.

PLAPT: Protein-Ligand Binding Affinity Prediction Using Pretrained Transformers

“…We begin by detailing the training, validation, and performance metrics on our test set, Benchmark1k2101. This is followed by a comparative analysis against state-of-the-art models on two independent benchmark datasets, Test-2016_290 and CSAR-HiQ_36, highlighting PLAPT's performance and generalizability [11,25]. [11], which recorded a validation RMSE (Root Mean Square Error) and MAE (Mean Absolute Error) of 1.338 and 1.034 respectively, PLAPT demonstrates significantly lower validation RMSE and MAE, meaning it is likely no-more overfit than the state-of-the-art.…”

“…PLAPT is designed to work solely with one-dimensional string inputs, necessitating just a protein sequence and the SMILES notation of a ligand for making predictions. This design contrasts with models such as CAPLA, which demand data on the protein pocket [11]. Refer to Figure 1 for an illustration of the inputs used by PLAPT.…”

“…Despite this, PLAPT's performance is more accurate than affinity_pred [18] on this benchmark, with a 9.10% improvement in RMSE and a 1.62% improvement in the MAE metric. Aside from affinity_pred, PLAPT surpasses CAPLA [11]…”

Rationalizing diverse binding mechanisms to the same protein fold: insights for ligand recognition and biosensor design

“…Indeed, aligning the known WIN conformer to its known Hbond acceptor geometry, followed by Rosetta FastDesign 42 , is sufficient to generate designs that recognize WIN with a nanomolar limit of detection. Developing new deep learning methods to learn and design small molecule-protein interactions is in vogue 15,[43][44][45][46][47][48] . We suggest that comparable attention should be placed with the choice of ligand conformer and rigid body orientation.…”

From Proteins to Ligands: Decoding Deep Learning Methods for Binding Affinity Prediction

“…Depending on the type of input data used during training, these deep learning (DL) methods can be broadly categorized as sequence- or complex-based methods . Complex-based methods − are trained on features from 3-dimensional (3D) protein–ligand complexes. Here we focus on sequence-based methods.…”