Analyzing effect of quadruple multiple sequence alignments on deep learning based protein inter-residue distance prediction

Abstract

Protein 3D structure prediction has advanced significantly in recent years due to improving contact prediction accuracy. This improvement has been largely due to deep learning approaches that predict inter-residue contacts and, more recently, distances using multiple sequence alignments (MSAs). In this work we present AttentiveDist, a novel approach that uses different MSAs generated with different E-values in a single model to increase the co-evolutionary information provided to the model. To determine the importance of each MSA’s feature at the inter-residue level, we added an attention layer to the deep neural network. We show that combining four MSAs of different E-value cutoffs improved the model prediction performance as compared to single E-value MSA features. A further improvement was observed when an attention layer was used and even more when additional prediction tasks of bond angle predictions were added. The improvement of distance predictions were successfully transferred to achieve better protein tertiary structure modeling.

LZerD Protein-Protein Docking Webserver Enhanced With de novo Structure Prediction

“…For structure modeling of individual proteins, the LZerD server uses AttentiveDist (Jain et al, 2021). If users have 3D structures of individual proteins to dock, they can skip the AttentiveDist step.…”

“…Although AttentiveDist was shown to have competitive performance at the time of the development (Jain et al, 2021), there are more recent methods that showed promising performance. Users are also encouraged to try such servers, perhaps those which performed well in recent CASP (Kryshtafovych et al, 2019).…”

“…Single-chain protein structure prediction methods have recently matured, and can often generate models in the absence of clear global template structures (Kryshtafovych et al, 2019). Such methods include our AttentiveDist (Jain et al, 2021), trRosetta (Yang et al, 2020), RaptorX (Xu and Wang, 2019), and QUARK (Zheng et al, 2019). Relative to the top existing servers participating in CASP13 (Kryshtafovych et al, 2019), AttentiveDist showed competitive performance when evaluated on the CASP13 (Kryshtafovych et al, 2019) dataset (Jain et al, 2021).…”

“…Such methods include our AttentiveDist (Jain et al, 2021), trRosetta (Yang et al, 2020), RaptorX (Xu and Wang, 2019), and QUARK (Zheng et al, 2019). Relative to the top existing servers participating in CASP13 (Kryshtafovych et al, 2019), AttentiveDist showed competitive performance when evaluated on the CASP13 (Kryshtafovych et al, 2019) dataset (Jain et al, 2021).…”

“…The underlying methods implemented in the LZerD webserver, LZerD (Christoffer et al, 2021), Multi-LZerD (Esquivel-Rodríguez et al, 2012), and AttentiveDist (Jain et al, 2021), and the ranksum model scoring function (Peterson et al, 2017a;Peterson et al, 2018a;Christoffer et al, 2020) have been rigorously examined in their original papers. In this article, we present the current version of the LZerD webserver, with new functionality of de novo prediction of subunit structures by AttentiveDist and applying symmetry constraints for homodimer modeling.…”