The small Molecule Drug Discovery field has been heavily dependent on suppressor discovery by structural binding prediction. Despite all successes in targeting different types of molecular targets in cells, many are considered undruggable. Over 85% of proteins and over 99% of RNAs are still considered hard to drug in cancer. The main challenge in suppressing their activity would be their unknown or super complicated structures. In addition, many of them present a dynamic 3D structure with no pocket. Since computational structural-based drug discovery has been developed for proteins with rigid structures and obvious pockets, it has not been successful in discovering small molecule candidate drugs against dynamic or unknown structures. In addition, structurally binding to some targets like RNAs does not guarantee their activity inhibition. Therefore, there has been a need for a computational approach to discover drugs against hard-to-drug targets with no structural information involved. I introduce a new small molecule drug discovery approach using Artificial Intelligence (AI), specifically Deep Learning (DL). This method does not require any input data from the sequence or the 3D structure of the target. Rather than targeting biomolecules' structure, AI models learn the biology of suppressing the target's activity called functional-based modeling. In three different projects, we prove the efficiency of AI-based functional-based drug discovery compared to traditional computational drug discovery. First, the collection of one of the biggest molecular datasets for molecular machine learning. MolData is one of the biggest categorized molecular datasets ever published for AI drug discovery. Second, I introduce RiboStrike, an AI-based model capable of discovering candidate drugs against micro RNAs. RiboStrike is the state-of-the-art model capable of discovering small molecule candidate drugs against RNA regardless of their size, structure, and coding functionality. We successfully discovered three candidate drugs against the functionality of miR21. Third, I introduce AMPdeep, an AI-based approach to discovering the hydrolysis of Anti-microbial Peptides (AMPs).
| Identifer | oai:union.ndltd.org:ucf.edu/oai:stars.library.ucf.edu:etd2020-2396 |
| Date | 01 January 2022 |
| Creators | Keshavarzi Arshadi, Arash |
| Publisher | STARS |
| Source Sets | University of Central Florida |
| Language | English |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Electronic Theses and Dissertations, 2020- |
Page generated in 0.0019 seconds