Talk2QSP

Agent-based framework that converts unstructured literature descriptions of clinical or preclinical scenarios into executable QSP/SBML model interventions, with dynamic human-in-the-loop ambiguity resolution.

Affiliation	Sanofi and bmedx (Sanofi)
First introduced	2026-05 (bioRxiv preprint v2)
Lifecycle stages	Experiment design
Autonomy level	Assistive (human-in-the-loop, dynamic HITL strategy)
Domain focus	Quantitative Systems Pharmacology (QSP) / kinetic ODE modelling
Availability	Closed (no public access announced in the preprint)

Approach

Combines semantic grounding of model entities, an LLM Scenario Extractor, and a dual-agent Scenario Mapper that issues discrete verifiable “work orders”. Humans resolve biological ambiguities interactively.

Validation

Seven subject-matter-expert-curated literature scenarios applied across four ODE/QSP models. All selected scenarios resolved into correct executable parameter changes, including multi-dose interventions, unit conversions, no-op scenarios, and ambiguity-triggered HITL cases.

Notable results

7/7 SME-curated scenarios resolved correctly. Outperformed standalone SBML-reasoning LLM calls and prior agentic efforts on processed SBML data (per preprint claim).

Primary paper

Kazemeini et al., “Talk2QSP: Deriving Executable Scenarios from Unstructured Literature via Human-in-the-Loop Agents,” bioRxiv 2026.05.06.723244.

Code

Not released.