SBIR-STTR Award

Adverse drug reactions (ADRs), more commonly known as drug side effects, are estimated to cause over 100,000 deaths in the US annually, are responsible for 6.5% of all hospital admissions, and 28% of clinical trial failures. Current pharmacological modeling efforts that are commonly used in the pharma industry (such as PK/PD) are generally empirical at the biomolecular level, meaning simplified mathematical terms are used to represent complex physiology. These modeling approaches are used to quantitatively understand therapeutic exposure-response relationships for clinical dosing, but have been less commonly applied to describe toxicological exposure-response relationships, with a few exceptions, as they lack the power to predict systemic cellular effects of pharmaceuticals that underlie ADRs. Elucidating the downstream and systemic effects of pharmaceuticals is critical to understanding ADR pathogenesis. Drugs can affect multiple proteins and each protein that they modulate may play roles in multiple cellular processes. Understanding this multi- factorial physiological response using systems biology methods will ultimately aid in better predicting ADRs before clinical trials using in vitro data. With the increasing emphasis on amassing large datasets, there is more and more publically available knowledge on pharmaceuticals, and their effects on cells, organs, and patients. However throughout all biomedical fields, analyzing complex datasets in a biologically coherent fashion has been a difficult challenge. The goal of this program is to develop a predictive computational platform analyzing gene-expression data sets from drug perturbed in vitro cell lines with metabolic and protein interaction networks for better understanding the systemic effects of over 700 approved pharmaceuticals with known ADRs. The platform, named ADR Predict, will use statistical machine learning approaches to identify network perturbation signatures that are highly predictive of specific ADRs. Developing ADR Predict has significant implications for the pharmaceutical industry. The initial commercialization opportunity of ADR Predict will is through service partnerships with pharmaceutical companies for accelerating and improving the drug development pipeline by mitigating risk of clinical trial safety failures. Further, this proposal will elucidate mechanisms of ADR pathogenesis that will be subsequently experimentally validated in Phase 2.

Public Health Relevance Statement:
Project Narrative Adverse drug reactions, more commonly known as drug side effects, cause over 100,000 deaths in the US annually, are responsible for 6.5% of all hospital admissions, and 28% of clinical trial failures. This proposal will develop a computational platform that can both predict drug side effects from preclinical data and elucidate associated biochemical mechanisms. The predictive platform will accelerate pharmaceutical partners' drug development pipelines by mitigating risk of safety failures during clinical trials of novel compounds.

Project Terms:
Admission activity; Adverse drug effect; Adverse effects; Adverse reactions; Affect; base; Benchmarking; Binding; Biochemical; Biochemical Pathway; Biological; Biological Markers; Cell Line; Cell physiology; cell type; Cells; cerivastatin; Cessation of life; Chemicals; Clinical; Clinical Trials; Clinical Trials Design; cohort; commercialization; Complex; cost; Data; Data Analyses; data integration; Data Set; Dose; drug development; Drug Industry; Drug toxicity; Economic Burden; Event; Failure; Gene Expression; Generations; Goals; Healthcare; Hospitals; Human; improved; In Vitro; Industry; Institutes; Intellectual Property; Knowledge; Link; Literature; Machine Learning; Marketing; Measures; Metabolic; Metabolic Pathway; Metabolism; Methods; Modeling; Names; network models; novel; Organ; Pathogenesis; Patient Noncompliance; Patients; Pharmaceutical Preparations; Pharmaceutical Services; Pharmacologic Substance; Phase; Physiological; Physiology; Play; pre-clinical; predictive modeling; Process; programs; Proteins; Reaction; Research; Research Infrastructure; response; Rewards; Risk; Rofecoxib; Role; Safety; Services; small molecule; Statistical Models; success; System; Systems Biology; Testing; Therapeutic; Time; Toxic effect; Withdrawal; Work

Adverse drug reactions (ADRs), more commonly known as drug side effects, are estimated to cause over 200,000 deaths in the US annually, are responsible for 6.5% of all hospital admissions, and 28% of clinical trial failures. ADRs are estimated to increase healthcare costs by $136 billion per year in the USA alone. Current safety and modeling efforts that are commonly used in the pharma industry (such as PK/PD) do not elucidate the complex pathophysiology underlying ADRs. These safety and modeling approaches are used predominantly to quantitatively understand exposure-response relationships for clinical dosing, but with a few exceptions do not focus on the cellular pharmacodynamic mechanisms of why drugs cause ADRs. Elucidating the downstream and systemic effects of pharmaceuticals is critical to understanding ADR pathogenesis and developing safer therapies. Drugs can affect multiple proteins and each protein that they modulate may play roles in multiple cellular processes. Systems biology and bioinformatics approaches coupled with machine learning are crucial for understanding the multi-factorial pathophysiology of ADRs. In Phase I of this program, we developed an in vitro transcriptomics based computational platform that 1) predicts drug-side effect liability equivalent to current gold-standard approaches that require considerably more information about the compound and its effects, 2) defines genes that are relevant to ADR pathophysiology, and 3) identifies therapeutically beneficial compounds for the ADR. Based on the computational platform, we discovered a repurposing opportunity for an off-patent, non-FDA approved drug in Parkinson’s Disease that we are currently pursuing towards clinical development. This drug significantly improves levodopa’s efficacy, without exacerbating the drug’s major side effect which often precludes levodopa’s use. In Phase II of this proposal, we will continue to develop and expand the ADR computational platform. Further, we will hone our focus on two key clinically and commercially relevant ADRs: antipsychotic induced tardive dyskinesia and radio-/chemo- therapy induced mucosal inflammation. We will generate rich datasets for these ADRs to both validate our in vitro platform with in vivo data and to understand the pathophysiology of these ADRs at an unprecedented level. Further, we will use the datasets to generate computational predictions for discovering/repurposing drugs to improve safety in psychiatric and cancer treatments. The best predictions will be subsequently tested in vitro and developed through partnerships and external funding mechanisms.

Public Health Relevance Statement:
Project Narrative Adverse drug reactions, more commonly known as drug side effects, cause over 200,000 deaths in the US annually, are responsible for 6.5% of all hospital admissions, and 28% of clinical trial failures. This proposal will continue development and deploy a computational platform that predict drug side effects' mechanisms for safer drug development and better patient outcomes. The predictive platform will be applied for improving cancer and psychiatric treatments.

Project Terms:
Adverse drug effect; adverse drug reaction; Adverse event; Affect; Algorithms; Animal Model; Antidepressive Agents; Antineoplastic Agents; Antipsychotic Agents; base; Biochemical; Bioinformatics; cancer therapy; Cell physiology; Cessation of life; chemoradiation; chemotherapy; Clinical; clinical development; Clinical Trials; commercialization; Complex; computational platform; Corpus striatum structure; Coupled; Data; data pipeline; Data Set; Databases; Development; Dose; drug development; drug discovery; Dyskinetic syndrome; Economic Burden; Etiology; Exposure to; Expression Profiling; Failure; Functional disorder; Funding Mechanisms; Gene Expression; Gene Expression Profile; Generations; Genes; Gold; Health Care Costs; Healthcare Systems; Hospitalization; Hospitals; improved; In Vitro; in vitro testing; in vivo; Industry; Infrastructure; Knowledge; Legal patent; Lesion; Levodopa; Literature; Machine Learning; metabolomics; Modeling; Mucositis; Mus; Neurons; novel therapeutics; off-patent; Parkinson Disease; Pathogenesis; Patient-Focused Outcomes; Pharmaceutical Preparations; Pharmacodynamics; pharmacokinetics and pharmacodynamics; Pharmacologic Substance; Pharmacology; pharmacovigilance; Phase; Play; prevent; programs; Proteins; Psychiatric therapeutic procedure; Radio; Reporting; response; Rodent; Rodent Model; Role; Safety; screening; Serious Adverse Event; side effect; Standardization; success; System; Systems Biology; Tardive Dyskinesia; Testing; Therapeutic; Therapeutic Uses; Tissues; transcriptome sequencing; transcriptomics; Validation