Explore Biological Phase Space Towards Desired Properties By In Silico Methods

Understanding the relationship between RNA sequence and function is important for revealing the functional roles and biomedical applications of RNA. The systematic evolution of ligands by exponential enrichment (SELEX) methodology enables an ef-fective search for functional aptamers, which discovers active aptamers for various tar-gets. To effectively explore sequence space and gain insight into the "black box" of SELEX experiments, we propose a computational aptamer selection approach, SELEX in silico. This approach encompasses several stages of sequence selection beginning from the comprehensive search of sequence space to identify sequences that are stable in a desired conformation, followed by rounds of molecular dynamic simulations of increasing accuracy. Six aptamers are predicted by SELEX in silico and experimental-ly verified to bind theophylline with high affinity-Kd ranging from 0.16μM to 0.52 μM while the aptamer with the best known affinity to theophylline has Kd=0.32μM. These results demonstrate the significant potential of SELEX in silico as a new method for aptamer sequence discovery and optimization. SELEX in silico method provided a practical computational solution to three key problems in aptamer sequence search, namely, how to design the initial library, how to open the "black box" of SELEX by an exhaustive sequence enrichment picture, and how to identify fitness peaks in a sequence space.Virtual library design, fast and accurate computational docking, subsequent drug design are three key fields of drug virtual screening. Fragment Optimized Growth (FOG) algorithm prepare virtual library with desired features, whose heart is a Markov Chain which adds fragments to the nascent molecule in a biased manner, depending on the frequency of specific fragment-fragment connections in the database of chemicals it was trained on. To examine practical performance of FOG, we perform virtual screen-ing for O-GlcNAc transferase inhibitors with FOG and ZINC library. It was found that FOG compounds have lower docking score and great structural diversity than ZINC at the last stage of virtual screening while ZINC compounds share several chemical fragments. By analyzing the docking conformation of the best 116 compounds hav-ing lowest docking scores, we design hundreds of novo compound by combining two fragments in the binding pocket. To evaluate the dynamic binding process and receptor flexibility, we adopted molecular dynamic simulation to simulate the docked binding conformation and chose the calculated the binding free energy as primary indicator for further experimental assay.By decomposing the docked conformations into separated chemical fragments, we visualized how fragments distributed in binding pocket. Surprisingly, certain type frag-ment agglomerate at certain position in the binding pocket which drove by the inter-action from particular residue environment of binding pocket. The agglomerated posi-tion for each chemical fragment were clustered and these position-based fragment were linked to form new compounds. Thus we proposed a drug design algorithm named Fragment Agglomeration Clustered Linkage (FACL), which aims to design adaptive compounds library based on virtual screening results and bridge high throughput vir-tual screening and fragment-based drug design. As for OGT inhibitors, Comound-1 which designed by FACL biased on ZINC virtual screening result have lower docking score,-14.59, than the best ZINC compound ZINC-32 (docking score=-11.49), while the docking score of UDP is-14.10.