PYTHIA: Deep Learning Approach for Local Protein Conformation Prediction

• Main Authors: Gabriel Cretin, Tatiana Galochkina, Alexandre G. de Brevern, Jean-Christophe Gelly
• Format: Article
• Language: English
• Published: MDPI AG 2021-08-01
• Series: International Journal of Molecular Sciences
• Subjects: protein structure; protein blocks; prediction; deep learning
• Online Access: https://www.mdpi.com/1422-0067/22/16/8831

Protein Blocks (PBs) are a widely used structural alphabet describing local protein backbone conformation in terms of 16 possible conformational states, adopted by five consecutive amino acids. The representation of complex protein 3D structures as 1D PB sequences was previously successfully applied to protein structure alignment and protein structure prediction. In the current study, we present a new model, PYTHIA (<span style="text-decoration: underline;"><b>p</b></span>redicting an<span style="text-decoration: underline;"><b>y</b></span> conforma<span style="text-decoration: underline;"><b>t</b></span>ion at <span style="text-decoration: underline;"><b>hi</b></span>gh <span style="text-decoration: underline;"><b>a</b></span>ccuracy), for the prediction of the protein local conformations in terms of PBs directly from the amino acid sequence. PYTHIA is based on a deep residual inception-inside-inception neural network with convolutional block attention modules, predicting 1 of 16 PB classes from evolutionary information combined to physicochemical properties of individual amino acids. PYTHIA clearly outperforms the LOCUSTRA reference method for all PB classes and demonstrates great performance for PB prediction on particularly challenging proteins from the CASP14 free modelling category.