Researchers Automated In Silico Cell Perturbation Screens
The ISP Platform uses Geneformer models to identify genetic sequences required for stem cell state transitions.
Updated on Sept. 24, 2026 in Biotech
Live Poll
Do you support prioritizing AI-driven virtual genetic screening to reduce reliance on laboratory animal testing?
Researchers have developed the ISP Platform, an automated system for in silico perturbation screens powered by the Geneformer deep learning model. The platform identifies optimal gene overexpression sequences to guide cellular state transitions across both human and mouse models.
Why it matters
This framework accelerates biological discovery by enabling precise in silico screens that identify candidate gene sequences for experimental validation. By streamlining this process, the platform offers a pathway to reduce exploratory animal use in developmental research.
The ISP Platform performed a targeted pairwise overexpression screen across 24 permutations of transcription factors. It successfully identified an optimal sequence for the primed-to-naive state transition in pluripotency.
The players
Geneformer
A transformer-based deep learning model designed for single-cell transcriptomics and predictive analysis of cellular gene networks.
The details
The platform automates perturbation screens by chaining each genetic intervention step from the preceding cell state, allowing for sequential multi-gene analysis. Users execute cross-species analysis by selecting specific Geneformer models—deep learning architectures trained on single-cell transcriptomics—to map homologous programs between human and mouse datasets. While human models were shown to capture mouse pluripotency programs, applying mouse-derived models to human cells revealed translation-related terms as key markers.
Timeline
September 24, 2026: The research findings were published.
The Tech Race
This platform extends the utility of the Geneformer model by providing a standardized, automated framework for interpreting gene expression perturbations. It represents a shift from static model analysis toward dynamic, sequential testing of cellular state transitions.
This research provides a computational tool for scientists to narrow down genetic targets before beginning costly laboratory experiments. It is currently a research-stage platform intended to refine experimental design rather than a clinical product.
The takeaway
The ISP Platform demonstrates a viable method for automating complex gene screening, setting a benchmark for future computational developmental studies. Researchers should monitor upcoming peer-reviewed validation results to see if these identified sequences hold true in vivo.
Further reading
For more on the latest research in cellular modeling, visit Biotech.
More information
Read the full scientific study paper on the bioRxiv server.
Source note: This article includes information reported by Biorxiv.
Live Poll
Do you support prioritizing AI-driven virtual genetic screening to reduce reliance on laboratory animal testing?
