Nova will be at ACoP 2026

Scaling nonlinear mixed-effects inference for QSP models: a mavoglurant case study

A surrogate-assisted SAEM framework that brings population parameter inference within reach of stiff, high-dimensional QSP models.

Specialized AI capabilities (‘agent skills’) to assist PBPK and QSP modeling and clinical trial simulations

An open-source collection of agent skills for PBPK and QSP work, keeping transparency and human expertise at the center of the workflow.

A Multi-Stage Calibration Framework for Digital Twins to Simulate Alternative ADC Dosing Regimens in NSCLC

Joint work with MSD, building patient-level mechanistic Digital Twins for an antibody-drug conjugate and using them to compare dosing regimens In Silico.

Efficient and Flexible Generation of Virtual Populations for Trial Simulation Using Simulated Annealing

Virtual population construction recast as combinatorial optimization: pick the sub-population that matches the target statistics directly.

Plot Digitization for Pharmacometrics: Extracting data from scientific plots using vision–language models, OCR, and axis calibration

A hybrid pipeline pairing vision language models with explicit axis calibration, so digitized series come back numerically grounded.

An Open Benchmark for LLM-Based reproduction of PBPK & QSP scientific publications

An open benchmark for evaluating how well LLM-assisted workflows rebuild a published PBPK or QSP model from the paper alone.

Also walked at the AI/ML SIG lunch poster walk, 12:00 PM – 1:30 PM in Chesapeake G-I.

A Mechanistic-Statistical QSP framework Integrating Clonal Competition and Survival Models to Optimize Androgen Deprivation Therapy in Prostate Cancer

Joint work with Ferring Pharmaceuticals, coupling clonal competition dynamics to a time-to-event model of metastasis and mortality.

A two-step model of individual patient progression-free survival in Mantle Cell Lymphoma

Individual-level tumor dynamics from the SHINE trial, linked to a per-patient progression-free survival distribution through an NLME and Cox pairing.

A mechanistic quantitative systems pharmacology model of atherosclerosis prospectively predicted LDL-C outcomes of the phase 3 CORALreef Lipids trial

Predictions were posted publicly before the readout. The week-24 LDL-C reduction came in at –54.5% predicted against –55.8% reported.

Accelerating QSP Calibration: A Hands-on Tutorial on Gaussian Process Surrogate Models and Stochastic Approximation Expectation-Maximization (SAEM)

A hands-on morning on training Gaussian process surrogates from simulation output, then running SAEM against them to calibrate complex models.