Effectiveness of iterative in silico–in vitro cycles

Determine whether iterative in silico–in vitro cycles of AI model training and experimental testing improve effectiveness over single-pass training on medium-throughput in vitro screens for target-specific small-molecule lead discovery using learning algorithms such as random forest and directed message passing neural networks.

Background

The paper discusses strategies for overcoming small labeled datasets in drug discovery, including generating target-specific training data via medium-throughput in vitro screens and training models such as random forests or directed message passing neural networks on these datasets. While these approaches have yielded fast and effective discovery of potential drug leads, the authors explicitly note uncertainty regarding the benefit of iterative feedback loops that alternate computational predictions with new experimental data collection.

References

The question remains, however, of whether iterative in silico-in vitro cycles would be even more effective.

— Data-centric challenges with the application and adoption of artificial intelligence for drug discovery  (2407.05150 - Ghislat et al., 2024) in Section 2.5 Small data

Third, the performance of the discovery system itself has not yet been established prospectively. The examples presented here demonstrate that CircuitATLAS can generate biologically plausible hypotheses and, for ATP1A3, nominate a target subsequently validated in vivo, but they do not establish its success rate, relative performance against expert-led or alternative computational approaches, or ability to predict therapeutic efficacy across indications. Several prospective, preregistered campaigns will therefore be required to determine whether accumulating experimental outcomes improve target prioritization and experimental efficiency over time.

— CircuitATLAS: Agentic reasoning over a systems neuroscience knowledge graph for target discovery in circuitopathies  (2610.09643 - Ocana-Santero et al., 7 Oct 2026) in Discussion, paragraph beginning “Several limitations remain”