As the infection of 2019-nCoV coronavirus is quickly developing into a global pneumonia epidemic, the careful analysis of its transmission and cellular mechanisms is sorely needed. In this Communication, we first analyzed two recent studies that concluded that snakes are the intermediate hosts of 2019-nCoV and that the 2019-nCoV spike protein insertions share a unique similarity to HIV-1. However, the reimplementation of the analyses, built on larger scale data sets using state-of-theart bioinformatics methods and databases, presents clear evidence that rebuts these conclusions. Next, using metagenomic samples from Manis javanica, we assembled a draft genome of the 2019-nCoV-like coronavirus, which shows 73% coverage and 91% sequence identity to the 2019-nCoV genome. In particular, the alignments of the spike surface glycoprotein receptor binding domain revealed four times more variations in the bat coronavirus RaTG13 than in the Manis coronavirus compared with 2019-nCoV, suggesting the pangolin as a missing link in the transmission of 2019-nCoV from bats to human.

SUMMARY Structure prediction for proteins lacking homologous templates in the Protein Data Bank (PDB) remains a significant unsolved problem. We developed a protocol, C-I-TASSER, to integrate interresidue contact maps from deep neural-network learning with the cutting-edge I-TASSER fragment assembly simulations. Large-scale benchmark tests showed that C-I-TASSER can fold more than twice the number of non-homologous proteins than the I-TASSER, which does not use contacts. When applied to a folding experiment on 8,266 unsolved Pfam families, C-I-TASSER successfully folded 4,162 domain families, including 504 folds that are not found in the PDB. Furthermore, it created correct folds for 85% of proteins in the SARS-CoV-2 genome, despite the quick mutation rate of the virus and sparse sequence profiles. The results demonstrated the critical importance of coupling whole-genome and metagenome-based evolutionary information with optimal structure assembly simulations for solving the problem of non-homologous protein structure prediction.

Comparison of ligand poses generated by protein–ligand docking programs has often been carried out with the assumption of direct atomic correspondence between ligand structures. However, this correspondence is not necessarily chemically relevant for symmetric molecules and can lead to an artificial inflation of ligand pose distance metrics, particularly those that depend on receptor superposition (rather than ligand superposition), such as docking root mean square deviation (RMSD). Several of the commonly-used RMSD calculation algorithms that correct for molecular symmetry do not take into account the bonding structure of molecules and can therefore result in non-physical atomic mapping. Here, we present DockRMSD, a docking pose distance calculator that converts the symmetry correction to a graph isomorphism searching problem, in which the optimal atomic mapping and RMSD calculation are performed by an exhaustive and fast matching search of all isomorphisms of the ligand structure graph. We show through evaluation of docking poses generated by AutoDock Vina on the CSAR Hi-Q set that DockRMSD is capable of deterministically identifying the minimum symmetry-corrected RMSD and is able to do so without significant loss of computational efficiency compared to other methods. The open-source DockRMSD program can be conveniently integrated with various docking pipelines to assist with accurate atomic mapping and RMSD calculations, which can therefore help improve docking performance, especially for ligand molecules with complicated structural symmetry.

We report the results of residue‐residue contact prediction of a new pipeline built purely on the learning of coevolutionary features in the CASP13 experiment. For a query sequence, the pipeline starts with the collection of multiple sequence alignments (MSAs) from multiple genome and metagenome sequence databases using two complementary Hidden Markov Model (HMM)‐based searching tools. Three profile matrices, built on covariance, precision, and pseudolikelihood maximization respectively, are then created from the MSAs, which are used as the input features of a deep residual convolutional neural network architecture for contact‐map training and prediction. Two ensembling strategies have been proposed to integrate the matrix features through end‐to‐end training and stacking, resulting in two complementary programs called TripletRes and ResTriplet, respectively. For the 31 free‐modeling domains that do not have homologous templates in the PDB, TripletRes and ResTriplet generated comparable results with an average accuracy of 0.640 and 0.646, respectively, for the top L/5 long‐range predictions, where 71% and 74% of the cases have an accuracy above 0.5. Detailed data analyses showed that the strength of the pipeline is due to the sensitive MSA construction and the advanced strategies for coevolutionary feature ensembling. Domain splitting was also found to help enhance the contact prediction performance. Nevertheless, contact models for tail regions, which often involve a high number of alignment gaps, and for targets with few homologous sequences are still suboptimal. Development of new approaches where the model is specifically trained on these regions and targets might help address these problems.