Introducing Our Perspective on Virtual Cells & the New TxPert Model
Learn how we're working to Predict, Explain, and Discover in drug discovery, and explore the capabilities of the new TxPert model.
Hi everyone,
We're excited to share some of the latest work from Valence Labs, Recursion's AI research engine, focusing on advancing drug discovery outcomes through cutting-edge computational methods. This past week, we've introduced a new perspective paper on our vision for building "virtual cells" and a state-of-the-art model, TxPert, our new tool for predicting transcriptional responses to genetic perturbations—representing an important step in this direction.
Our Framework: Predict – Explain – Discover
Our research is guided by a framework we call Predict—Explain—Discover. This involves developing AI systems that can:
Predict the functional response of cells to perturbations across diverse biological contexts, timepoints, and modalities.
Explain these responses by identifying key biomolecular interactions, causal pathways, and context-dependent regulatory mechanisms.
Discover novel biological insights and actionable therapeutic hypotheses through iterative, lab-in-the-loop experimentation.
These principles guide our progress toward virtual cells: mechanistic models of functional response that can enable therapeutic discovery. We share our views on this new frontier in AI-enabled drug discovery, built on our experience at Recursion.
Introducing TxPert: Predicting Transcriptional Responses
As an important first step toward the Predict capability of our framework, we are also introducing TxPert: a state-of-the-art model for predicting transcriptional responses to genetic perturbations, especially in out-of-distribution scenarios.
Observing how cells respond to genetic changes is central to understanding disease and designing more effective therapies. TxPert takes a unifying approach to address this challenge by leveraging graph neural networks and multiple biological knowledge graphs, including curated public resources and Recursion's unique large-scale relationship maps. This integration provides TxPert with a structured understanding of biological relationships, enabling it to generalize across unseen single-gene perturbations, novel combinations of gene perturbations, and even new cell types.
To make TxPert accessible to researchers, we've also launched the TxPert App. This hands-on tool allows you to:
Select a cell type and gene to perturb
Visualize local and extended gene interaction networks
Analyze predicted expression changes in observed genes
Explore TxPert further:
Together, this work represents an important step towards our mission of decoding biology to radically improve lives. We are excited to share our vision and invite the scientific community to explore these resources together.
Reminder: MoML 2025: Poster Submissions Still Open!
MoML 2025 is happening June 18 at Mila, and we’d love to see you there. It’s a great opportunity to connect with researchers at the intersection of machine learning and molecular science—and to showcase your work!
Submit a poster for MoML: [Here]
Register for MoML: [Here]
See you soon,
The Portal Team
 | A guest post by
|
 | A guest post by
|