AlphaFold Blindness to Topological Barriers Affects Its Ability to Correctly Predict Proteins’ Topology

Abstract

AlphaFold is a groundbreaking deep learning tool for protein structure prediction. It achieved remarkable accuracy in modeling many 3D structures while taking as the user input only the known amino acid sequence of proteins in question. Intriguingly though, in the early steps of each individual structure prediction procedure, AlphaFold does not respect topological barriers that, in real proteins, result from the reciprocal impermeability of polypeptide chains. This study aims to investigate how this failure to respect topological barriers affects AlphaFold predictions with respect to the topology of protein chains. We focus on such classes of proteins that, during their natural folding, reproducibly form the same knot type on their linear polypeptide chain, as revealed by their crystallographic analysis. We use partially artificial test constructs in which the mutual non-permeability of polypeptide chains should not permit the formation of complex composite knots during natural protein folding. We find that despite the formal impossibility that the protein folding process could produce such knots, AlphaFold predicts these proteins to form complex composite knots. Our study underscores the necessity for cautious interpretation and further validation of topological features in protein structures predicted by AlphaFold.

A Comprehensive Analysis of Auxin Response Factor Gene Family in Melastoma dodecandrum Genome

“…Additionally, the WoLF PSORT tool (https://wolfpsort.hgc.jp/, accessed on 17 October 2023) was utilized to forecast the cellular localization of these proteins [55]. Moreover, for the prediction of the tertiary structures of the MedARFs, we utilized the Protein Structure Prediction tools (AlphaFold2.3.2) (https://colab.research.google.com/github/ deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb, accessed on 16 November 2023) [56]. In addition, we used SOMPA (https://npsa-pbil.ibcp.fr/cgi-bin/npsa_automat.…”

Knotted artifacts in predicted 3D RNA structures

“…This topological energy barrier hinders the formation of knotted proteins. Since methods that predict the native state of biopolymers do not fold molecules de novo but directly guess these native states, they have a harder job avoiding states unavailable due to topological energy barriers [ 39 , 40 ].…”