Improving Drug Development by Connecting Medicinal Chemistry with Drug

Repositioning and Modern Machine Learning Methods

GRAND Seminar May 1, noon, Tuesday, 2012, ENGR 4201

Iwona E. Weidlich
UMBC

Host:

Huzefa Rangwala

Abstract:

Developing drug candidates from scratch has turned into a billion-dollar expense that is not delivering enough profitable products to market. Novel approaches which merge chemistry with biology and informatics contribute to the development of selective lifesaving drugs needed by patients. We implement machine learning classifiers for HTS Data Analysis, Screening and drug repurposing with high probability of selecting drug candidates eligible for Phase II of clinical study free from ADME/Tox-related problems. We used small molecule bioactivity data for HCV RNA Polymerase to train and test QSAR models and apply these robust models for compound ranking and hit identification in drug repositioning techniques. Random Forest and kNN algorithms were used with Morgan fingerprints of 679 small molecules with curated IC50 values. After filtering various drug-like databases (DrugBank, MDL, NIAID-NIH, ComGenex) compounds were selected and tested against HCV. We discuss the challenges in drug repositioning faced in academia, government and pharmaceutical industry.

Short Bio:

Dr. Weidlich received her Ph.D. in Pharmaceutical Sciences from the University of Medical Sciences, Poznan, Poland in 2005. Her Ph.D. research focused on developing anti-cancer agents that are designed to be activated only inside a cancerous cell but have benign form in the systemic circulation. Her research interests also include using very large collections of chemical databases to filter and extract relevant subsets of molecules for closer analysis, and performing physicochemical and ADME/Tox property predictions. Specifically, she is interested in designing and evaluation of novel anti-cancer agents. She joined the Computer-Aided Drug Design (CADD) group at the Chemical Biology Laboratory, National Cancer Institute in Frederick, NIH as a postdoctoral fellow in October 2005. Dr. Weidlich has been conducting in silico screening for the inhibitors of cancer DNA, specifically tyrosyl-DNA phosphodiesterase (Tdp1) and Shc Src homology 2 (SH2) domain. She designed new, more powerful Tdp1 inhibitors and Shc SH2 domain-binding inhibitors: tetramer peptide-peptoid hybrids exhibiting up to 40-fold increase in affinity.

She accepted a faculty position at the University of Maryland Baltimore County (UMBC) in October 2010, remaining affiliated with NCI/NIH as a Guest Researcher. At UMBC she employs virtual screening to identify novel allosteric inhibitors of the Hepatitis C Virus NS5B Polymerase and also seeks to understand the mechanism by which small molecules inhibit NS5B.

Her current research is also related to combining robust QSAR modeling with drug repositioning and systems biology. This project of hers focuses on resolving problems arising from the management and analysis of huge amount of biological data as well as detailed study of multi-domain proteins and their function. Dr. Weidlich proposes techniques which could result in successful drug selection, and re-investigation of existing drugs for new therapeutic indications.