Development of Predictive Algorithms for In Silico Drug Toxicity and Efficacy Assessment

Period of Performance: 10/09/2002 - 06/10/2003


Phase 1 SBIR

Recipient Firm

Quantum Intelligence, Inc.
3375 Scott Blvd, Suite 100
Santa Clara, CA 95054
Principal Investigator


The aim and purpose of this proposal is to research and build a Quantum Intelligence System (QIS) and demonstrate its technical innovation and commercial feasibility for predicting in silico drug toxicity and efficacy. QIS integrates a collection of state-of-the-art data mining and optimization technologies. QIS first employs the best practices in database and data warehousing area to integrate the important information from diversified databases, then employs innovative data mining techniques which can are specially designed to handle large dimensional numeric and text data jointly to analyze, classify and predict the potential toxicity and efficacy of drug candidates. The effort is critical and significant because QIS technology make it possible to look at heterogeneous databases as a whole and study all the aspects of a drug discovery process jointly. So it is very likely that some previously unknown insight that could come out and lead to significant breakthrough. If successful, the resultant QIS can be applied to military defense interest of chemical and biological agents and pathogens. QIS should be also readily applicable to civil efforts of drug discovery. QIS would help drug candidates to "fail fast" before expensive later-phase trials, thus saving time and money. Making effective use of all of the existing, valuable data and literature is critical for a breakthrough in the area. Quantum Intelligence System, designed on the basis of large data warehousing and data mining, is able to look for patterns and build predictive models across multiple and heterogeneous databases. We believe that only by looking at all the data jointly, the patterns and actionable information that previously unknown will emerge and predictive accuracy can be greatly improved QIS has advantages over other data mining as follows · QIS focuses on an integrated knowledge source for drug candidates. · QIS predicts and classifies drug toxicity and efficacy from the multidimensional data by using our innovative Projection Pursuit Learning (PPLN), Knowledge Space Link Analysis (KPLA) and Structured Dynamic Programming (SDP) technologies. · QIS Knowledge Space Link Analysis is able to analyze both numerical data and text data. Descriptive features can be extracted from science literature and used them for understanding relationship between drug candidates. Other Military Applications: Analyze terrorists' behavior, discover potential terror threats; automatically allocate operation workload for undersea warfare; data fusion and sensor management for adaptive flight control. Supply Chain: Inventory management, production planning and control, resource allocation, reliability, maintenance, material handling, logistics, distribution of a supply chain. Manufacturing: Parsing existing work-stream data for chip production flow and optimizing the microchips' yield while minimizing costs. Health Care: Discover clinical activity patterns for automatic early detection, prevention and best practice. Bioinformatics: Find genes that are most responsible for a biology process to answer how gene expression is switched on and off in the context of each other during the course of a biologic process. Business Intelligence: Generic tools for revenue and resource management, customer profiling, targeted marketing, optimizing store location and store layout, customer attrition, customer win-back and cross-channel customer migration, capital budget allocation and risk management.