Biology subjects

Nussinov, R.

Publications and source records attributed to Nussinov, R..

21 records · Page 2Linked to original sources

bioRxiv2020

A deep learning framework for elucidating whole-genome chemical interaction space

Molecular interaction is the foundation of biological process. Elucidation of genome-wide binding partners of a biomolecule will address many questions in biomedicine. However, ligands of a vast number of proteins remain elusive. Existing methods mostly fail when the protein of interest is dissimilar from those with known functions or structures. We develop a new deep learning framework DISAE that incorporates biological knowledge into self-supervised learning techniques for predicting ligands of novel unannotated proteins on a genome-scale. In the rigorous benchmark studies, DISAE outperforms state-of-the-art methods by a significant margin. The interpretability analysis of DISAE suggests that it learns biologically meaningful information. We further use DISAE to assign ligands to human orphan G-Protein Coupled Receptors (GPCRs) and to cluster the human GPCRome by integrating their phylogenetic and ligand relationships. The promising results of DISAE open an avenue for exploring the chemical landscape of entire sequenced genomes.

bioinformatics↗

bioRxiv2020

The structural basis of the oncogenic mutant K-Ras4B homodimers

Ras proteins activate their effectors through physical interactions in response to the various extracellular stimuli at the plasma membrane. Oncogenic Ras forms dimer and nanoclusters at the plasma membrane, boosting the downstream MAPK signal. It was reported that K-Ras4B can dimerize through two major interfaces: (i) the effector lobe interface, mapped to Switch I and effector binding regions; (ii) the allosteric lobe interface involving 3 and 4 helices. Recent experiments showed that constitutively active, oncogenic mutant K-Ras4BG12D dimers are enriched in the plasma membrane. Here, we perform molecular dynamics simulations of K-Ras4BG12D homodimers aiming to quantify the two major interfaces in atomic level. To examine the effect of mutations on dimerization, two double mutations, K101D/R102E on the allosteric lobe and R41E/K42D on the effector lobe interfaces were added to the K-Ras4BG12D dimer simulations. We observed that the effector lobe K-Ras4BG12D dimer is stable, while the allosteric lobe dimer alters its helical interface during the simulations, presenting multiple conformations. The K101D/R102E mutations slightly weakens the allosteric lobe interface. However, the R41E/K42D mutations disrupt the effector lobe interface. Using the homo-oligomers prediction server, we obtained trimeric, tetrameric, and pentameric complexes with the allosteric lobe K-Ras4BG12D dimers. However, the allosteric lobe dimer with the K101D/R102E mutations is not capable of generating multiple higher order structures. Our detailed interface analysis may help to develop inhibitor design targeting functional Ras dimerization and high order oligomerization at the membrane signaling platform.

biophysics↗

bioRxiv2020

Genome-wide Prediction of Small Molecule Binding to Remote Orphan Proteins Using Distilled Sequence Alignment Embedding

Endogenous or surrogate ligands of a vast number of proteins remain unknown. Identification of small molecules that bind to these orphan proteins will not only shed new light into their biological functions but also provide new opportunities for drug discovery. Deep learning plays an increasing role in the prediction of chemical-protein interactions, but it faces several challenges in protein deorphanization. Bioassay data are highly biased to certain proteins, making it difficult to train a generalizable machine learning model for the proteins that are dissimilar from the ones in the training data set. Pre-training offers a general solution to improving the model generalization, but needs incorporation of domain knowledge and customization of task-specific supervised learning. To address these challenges, we develop a novel protein pre-training method, DIstilled Sequence Alignment Embedding (DISAE), and a module-based fine-tuning strategy for the protein deorphanization. In the benchmark studies, DISAE significantly improves the generalizability and outperforms the state-of-the-art methods with a large margin. The interpretability analysis of pre-trained model suggests that it learns biologically meaningful information. We further use DISAE to assign ligands to 649 human orphan G-Protein Coupled Receptors (GPCRs) and to cluster the human GPCRome by integrating their phylogenetic and ligand relationships. The promising results of DISAE open an avenue for exploring the chemical landscape of entire sequenced genomes.

systems biology↗