Improving peptide-protein docking with AlphaFold-Multimer using forced sampling

Abstract

Protein interactions are key in vital biological processes. In many cases, particularly in regulation, this interaction is between a protein and a shorter peptide fragment. Such peptides are often part of larger disordered regions in other proteins. The flexible nature of peptides enables the rapid yet specific regulation of important functions in cells, such as their life cycle. Consequently, knowledge of the molecular details of peptide-protein interactions is crucial for understanding and altering their function, and many specialized computational methods have been developed to study them. The recent release of AlphaFold and AlphaFold-Multimer has led to a leap in accuracy for the computational modeling of proteins. In this study, the ability of AlphaFold to predict which peptides and proteins interact, as well as its accuracy in modeling the resulting interaction complexes, are benchmarked against established methods. We find that AlphaFold-Multimer predicts the structure of peptide-protein complexes with acceptable or better quality (DockQ ≥0.23) for 66 of the 112 complexes investigated—25 of which were high quality (DockQ ≥0.8). This is a massive improvement on previous methods with 23 or 47 acceptable models and only four or eight high quality models, when using energy-based docking or interaction templates, respectively. In addition, AlphaFold-Multimer can be used to predict whether a peptide and a protein will interact. At 1% false positives, AlphaFold-Multimer found 26% of the possible interactions with a precision of 85%, the best among the methods benchmarked. However, the most interesting result is the possibility of improving AlphaFold by randomly perturbing the neural network weights to force the network to sample more of the conformational space. This increases the number of acceptable models from 66 to 75 and improves the median DockQ from 0.47 to 0.55 (17%) for first ranked models. The best possible DockQ improves from 0.58 to 0.72 (24%), indicating that selecting the best possible model is still a challenge. This scheme of generating more structures with AlphaFold should be generally useful for many applications involving multiple states, flexible regions, and disorder.

Computational Approaches Drive Developments in Immune-Oncology Therapies for PD-1/PD-L1 Immune Checkpoint Inhibitors

“…They concluded that this assay had a better prediction score for identifying stronger peptide binders that also adopt stable secondary structures upon binding [ 92 ]. Johansson-Åkhe et al [ 93 ] also show the importance of AlphaFold, in this case, AlphaFold-Multimer. AlphaFold-Multimer is shown to be capable of predicting the structure of peptide–protein complexes with acceptable or better accuracy than previous models and also has the capacity to predict whether a peptide and a protein will interact, thus improving peptide–protein docking [ 93 ].…”

“…One of the most recently developed computational methods, AlphaFold, has revolutionized structural biology by predicting highly accurate structures of proteins and their complexes with peptides, antibodies and proteins [ 92 , 93 , 94 ]. In addition to that, AlphaFold can also be useful for protein–peptide systems and drug discovery, as it can help identify the highest affinity binder among a set of peptides.…”

Evolution of the ribbon-like organization of the Golgi apparatus in animal cells

“…Models of complexes between conspecific GRASP/Golgin-45 pairs were built with the Colab implementation of AlphaFold2 77 , using MMseqs2 to generate multiple sequences alignments 90 . To obtain reliable predictions of the protein-peptide complexes, AlphaFold-Multimer version v2 was used, with 12 recycles for the generation of each model 91 . Complexes were built without the use of structural templates and without Amber refinement as this step does not introduce substantial improvement, while significantly increasing computational time.…”

“…90 . To obtain reliable predictions of the protein-peptide complexes, AlphaFold-Multimer version v2 was used, with 12 recycles for the generation of each model 91 . Complexes were built without the use of structural templates and without Amber refinement as this step does not introduce substantial improvement, while significantly increasing computational time.…”

“…Recent benchmark studies have demonstrated that the predictive power of AF2 extends beyond the production of the mere structural models, yielding accurate results also for protein-protein and protein-peptide complexes, even when they imply conformational changes, and providing reliable hints on the effect of missense mutations S17 . In the case of protein-peptide complexes, models with higher confidence can be obtained by increasing the number of recycles during model generation, provided that a sufficient number of sequences are detected during the generation of the multiple sequence alignments (MSAs) S18,19 . In general, AF2 predicted models are evaluated and ranked based on a per residue score, the predicted local distance difference test (pLDDT).…”