We developed interpretable machine learning-based network technologies to study protein-protein/protein-RNA binding dynamics, enhancing sample efficiency and magnifying interaction changes. The strategy has been successfully applied to study multi-body interactions in complex biomolecular systems. We successfully developed an interpretable machine learning-based network RNet. RNetsite employs a machine learning-based network decomposition algorithm to predict RNA binding sites by analyzing the local and global network properties. Our study shows that RNetsite outperforms existing methods, achieving precision values as high as 0.701 on TE18 and 0.788 on RB9 tests. We also developed RNetdyn, a distance-based dynamic graph algorithm, to characterize the consequences of the interface dynamical behavior upon inhibitor binding. The simulation testing of competitive inhibitors indicates that RNetdyn outperforms the traditional method by 30% (Briefings in Bioinformatics, 2024. Highlighted story). Tcf1- and Lef1-deficient CD8+ T cells exhibit histone hyperacetylation, ascribed to intrinsic histone deacetylase (HDAC) activity in Tcf1 and Lef1. Mutation of five conserved amino acids in the Tcf1 HDAC domain diminishes HDAC activity and the ability to suppress CD4+ lineage genes in CD8+ T cells (Nature Immunology, 2016. Highlighted story).

We developed automated and efficient computational tools to address RNA structural challenges. One successful example is DRPScore, a deep-learning-based approach to identifying native-like RNA-protein structures. DRPScore is tested on representative sets of RNA-protein complexes of various degrees of binding-induced conformation change ranging from fully rigid docking (bound-bound) to fully flexible docking (unbound-unbound). Out of the top 20 predictions, DRPScore consistently outperforms the existing methods with a roughly 10.53-15.79% improvement, even for the most difficult unbound cases (Nature Communications, 2023). We also developed efficient tools (3dRNA, 3dRNAscore, DIRECT, Rbind, and RPDescriptor) to accurately build the RNA tertiary structure from the smallest secondary elements (SSEs) using sequences, secondary structure, and interaction information. It takes advantage of both the atom-level precision of the structure and the nucleotide-level tertiary interactions. Our developed RNA structural computational tools are easy to use and are increasingly used to solve structure-related problems in the RNA community.

We developed the Physics+AI-based computational approaches and applied them to drug design. First, we proposed computational tools (Rpflex, RPpocket, RPocket, and HKPocket) for pocket detection and topology calculation. By analyzing the sequence conservation, critical interactions, and hydrophobicity of identified drug pockets, potential inhibitors can be selected and optimized. Then, we applied the developed tools for human disease-related study. HIV protein Tat serves as a mediator to bring host elongation protein P-TEFb onto the viral TAR RNA for active transcription. Our work can aid new experimental approaches to exploit the synergetic effect of combined inhibitors (F07#13 and JB181) in removing Tat from both sides for possibly more efficient Tat degradation, which can set HIV-infected cells into deep latency (Biophysical Journal, 2021. Cover story). We proposed one dynamic geometry design approach for RNA inhibitors with only a tiny pool of designed geometrically compatible scaffold candidates. First, our method uses graph-based tree decomposition to explore the complementarity rigid binding cyclic peptide and design the amino acid side chain length and charge to fit the RNA pocket. Then, we perform an energy-based dynamical network algorithm to optimize the inhibitor-RNA hydrogen bonds. Dynamic geometry-guided design yields successful inhibitors with low micromolar binding affinity scaffolds and experimentally competes with the natural RNA chaperone (Physical Chemistry Chemical Physics 2023. Cover story).