A Novel Cognitive Knowledge Harvesting Approach to Train a Self Learning System for Drug Predictive Models
CSEF · 2016 Mathematics & Computer Science
Overview
Objectives/Goals The goal is to develop an incremental self-learning model that helps in the prediction of drug behavior effectively increasing the net knowledge without human intervention Methods/Materials I built an iterative learning model that makes measurable progress with each iteration. I designed a BioEntity based feature-vector abstraction for capturing multidimensional harvested interactions. It taps into ontological specificity across knowledge networks to exploit existing knowledge, and explore new knowledge. Results My knowledge network generated unprecedented novel cross validations that include 1.16 million BioEntities, 2.008 million relationships, and 6.578m properties. The final knowledge sets were generated as a result of successive approximations of incremental knowledge gain/loss and the number of iterations have a finite upper bound that converges to an order of O(N*LogN) in terms of cost complexity. Through each iteration, new knowledge insights were generated that asymptotically and monotonically converge to the expected knowledge set with an 81% accuracy level. The P-value of the confidence aggregate score of the interactions consensus across 30 sample runs was 0.9236 at 0.05 significance level. Conclusions/Discussion A set of Bio-Entities is the initial input which represents a specific assay and constitutes the research problem of interest. This assay, when studied for harvesting knowledge generation created a meaningful knowledge set that more closely aligns with an associated confidence score. The generated knowledge was corroborated with a Lapatinib use case subgraph that included EGFR/HER2 receptor binding, MAPK signaling pathways reported in literature.The project successfully demonstrates an operational self learning system that can evolve, update knowledge continuously, and improve drug behavior predictions in randomized cell# based assays. Knowledge was corroborated with a Lapatinib use case sub graph that included EGFR/HER2 receptor binding, MAPK signaling pathways reported in literature.
Summary statement
Developing iterative self-learning knowledge network topologies from curated life science datasets which converge to expected knowledge, and performing ontological feature based analysis using this knowledge to help predict drug behavior.
Help received
Archana Gangakhedka, Drug Researcher & Dr. Lato provided verification of experimental methodology, validation of results, tips for presentation. My teacher Mrs. Taylor sponsored my project.
Competition history
- CSEF 2016
Resources
Related projects
ISEF · 2016
Self Driving Pharma: A Novel Cognitive Knowledge Harvesting Approach to Train a Self-Learning System For Drug Predictive Models through Multi-Dimensional Bio-Entity Feature-Vector Based Topological Data Analyses
ISEF · 2023
Using Artificial Intelligence Neural Networks to Repurpose Pre-Existing Drugs and Analyze Drug Relationships, Year 3
CSEF · 2015
A Systems Biology Approach to Optimize Prediction of Efficacy of Pathway Targeted Therapies in Cancer
ISEF · 2022
Using Artificial Intelligence Graph Neural Networks To Repurpose Pre-existing Drugs and Analyze Drug Relationships: Year 2
ISEF · 2023
Integrating Machine Learning with 3D Organoid Modeling to Identify Biomarkers to Combat Drug Resistance in Cancer
CSEF · 2014
Development of a Novel Computer-Aided Framework for the Discovery of Synergistic Chemotherapy Combinations
ISEF · 2022
Novel Reinforcement Learning Framework for Synthesizable Drug Optimization
ISEF · 2021
Novel Evolutionary Artificial Intelligence Methods for De Novo Drug Design
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: California Science & Engineering Fair public projects