Complementing QSP with Machine Learning to Predict IGA-Based Population Response in an Atopic Dermatitis Model

Authors: Mittal A, Parish N, Simpson DA, Suderman R
Conference: ACoP
Keywords: ACoP 2026, Atopic Dermatitis (AD), Atopic Dermatitis Model, machine learning, qsp, QSP (quantitative systems pharmacology)

Objective

  • Atopic Dermatitis (AD) is a chronic skin condition characterized by inflammation, itching, and redness. AD is an active area of drug development, with 120+ drug candidates in various stages of pre-clinical and clinical development[1].
  • Quantitative Systems Pharmacology (QSP) combines mechanistic disease understanding with drug pharmacodynamics and can guide dose selection and trial design in AD.
  • AD clinical trials differ in their choice of endpoints, which complicates trial design and limits comparability across studies[2]. A QSP model capable of predicting multiple endpoints could enable cross-therapeutic comparison.
  • The Investigator’s Global Assessment (IGA) is a clinician-administered composite score (typical range 0-4) designed to quantify disease severity. The percentage of patients achieving IGA 0 or 1 serves as a key endpoint in trials enrolling moderate-to-severe AD patients. IGA is inherently subjective and the lack of published data directly linking IGA to disease biomarkers presents challenges for mechanistic modeling.
  • We implement a machine learning strategy to overcome these limitations and integrate a population-level IGA response endpoint into an established AD QSP model.

By Anjali Mittal, Nicole Parish, Destiny A. Simpson, Ryan Suderman

ACoP 2026, October 11-14, 2026, National Harbor, Maryland