What Valence Labs does

Valence Labs is the artificial intelligence research engine of Recursion (NASDAQ: RECR), operating as a wholly-owned subsidiary headquartered in Montreal, Quebec, with additional research presence in Salt Lake City, Utah and Vancouver (via the University of British Columbia). Founded in 2018 and employing roughly 56 people, the company's stated purpose is to develop computational methods that predict, explain, and decode biology, organized around a "Virtual Cells" framework with three pillars: predicting cellular responses to perturbations, generating causal explanations for molecular interventions, and operating lab-in-the-loop engines for therapeutic hypothesis generation. Primary technical outputs include TxPert, a graph-neural-network model for predicting transcriptional responses to single and combinatorial genetic perturbations; OpenQDC, an open-source repository consolidating more than 40 quantum mechanics datasets covering roughly 1.5 billion molecular geometries; the Polaris Hub open benchmarking platform; the RxRx3-core reference dataset; and the Valence Portal community site.

The company's technology stack is built on Recursion's proprietary interventional biology and chemistry dataset (reported at over 65 petabytes), phenomics and transcriptomics foundation models, custom GNN architectures (Exphormer-MG, GAT-MultiLayer) integrated with biological knowledge graphs, and BioHive-2, a supercomputer developed in partnership with NVIDIA and described as the most powerful in biopharma. Valence releases its research as peer-reviewed publications, open-source software on GitHub, and open Python libraries distributed via pip and conda, and it employs a community-led go-to-market model aimed at academic and pharmaceutical researchers rather than direct commercial sales.

Valence Labs does not operate an independent commercial product line or pricing model; its outputs feed directly into Recursion's drug discovery pipeline and platform strategy. Revenue, where it exists at all, is captured at the parent level. Customer and partner relationships are concentrated: the effective principal "customer" is Recursion itself, while named external collaborators include NVIDIA (core infrastructure), Google Cloud (distribution of Recursion foundation models), Tempus (patient-level data collaboration), and the University of British Columbia (academic co-authorship). The company has no independent venture funding, no public ticker, and no segment-level revenue disclosure in the available source material.

Name

Valence Labs

Legal name

Valence Discovery Inc.

Website

https://valencelabs.ca

Company type

Private

Operating status

Operating

Headcount range

51–100 employees

Short description

Valence Labs is the wholly-owned AI research subsidiary of Recursion, developing virtual-cell models that predict and explain cellular responses to genetic and chemical perturbations using proprietary multimodal datasets, custom graph neural networks, and the BioHive-2 supercomputer, with outputs released as open-source tools and peer-reviewed publications for the academic and pharmaceutical research community.

Ownership category

akta.pro rank

Product category

AI Drug Discovery

NAICS

Research and Development in Biotechnology (except Nanobiotechnology) (541714), Scientific Research and Development Services (5417)

SIC

Services-Commercial Physical & Biological Research (8731)

akta.pro primary industry

AI/ML Platforms for Drug Discovery & Experiment Optimization (HLAGAJAO)

akta.pro secondary industries

Scientific Data Management & Data Platforms (SDMS/Data Lakehouse for R&D) (HLAGAJAC), Cheminformatics & Molecular Modeling / CADD (HLAGAJAE)

Keywords

• AI drug discovery
• Foundation models biology
• Perturbation prediction models
• Open-source quantum datasets
• Virtual cell modeling

Headquarters

HQ city

Montréal

HQ country

Canada

HQ region

North America

Offices3 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Infrastructure, Operations

Revenue model

1. Research & Development (Parent Company): Valence Labs operates as Recursion's AI research engine and does not appear to have independent revenue streams. The company develops new AI methods for predicting, explaining, and decoding biology to support Recursion's drug discovery platform and pipeline.

Go-to-market motion1 record

Distribution channels3 records

Marketing channels6 records

Core offering

Valence Labs is Recursion's AI research engine, developing new ways to predict, explain, and decode biology using AI and machine learning. It builds open-source computational models (TxPert, OpenQDC) and the Virtual Cells framework for predicting cellular responses to genetic and chemical perturbations, targeting drug discovery applications. The company leverages Recursion's 65+ petabyte proprietary biological dataset, automated labs, and BioHive-2 supercomputer to produce research outputs distributed through open-source channels, publications, and community platforms.

Product overview

Valence Labs is Recursion's AI research engine focused on decoding biology through computational modeling. The company develops virtual cell systems — mechanistic models of cellular function — organized around three pillars: Predict (modeling functional responses to perturbations at unprecedented scale), Explain (generating causal explanations for molecular interventions), and Discover (lab-in-the-loop engines for iterative therapeutic hypothesis generation). The product portfolio includes TxPert (a state-of-the-art transcriptomic perturbation prediction model), OpenQDC (an open-source quantum mechanics dataset hub), Polaris Hub (an open benchmarking platform), RxRx3-core (a benchmark dataset), and the Valence Portal (a research community platform), all powered by Recursion's automated biology/chemistry labs and BioHive supercomputing infrastructure.

Differentiator

Problem solved

Functional benefit

Brands

• TxPert: State-of-the-art model for predicting transcriptional responses to genetic perturbations
• OpenQDC
• Virtual Cells
• Polaris

Products and services

• TxPert A state-of-the-art deep learning model that predicts transcriptional responses to genetic perturbations (single and combinatorial) across cell types. Uses graph neural networks integrated with biological knowledge graphs (STRINGdb, Gene Ontology, PxMap, TxMap) to achieve robust out-of-distribution generalization for unseen gene knockdowns, perturbation combinations, and new cellular contexts.
• OpenQDC An open-source hub of machine-learning-ready quantum mechanics datasets, consolidating 40+ QM datasets covering 1.5 billion geometries across 70 atom species and 250+ QM methods. Delivered as a Python library (pip install openqdc) with standardized metadata, units, and normalization for MLIP (machine learning interatomic potential) research.
• Virtual Cells A conceptual framework and computational platform for mechanistic models of cellular function built on three pillars: Predict (accurate prediction of cellular responses to perturbations), Explain (causal explanations for molecular interventions), and Discover (lab-in-the-loop engines for iterative therapeutic hypothesis generation and refinement).
• Polaris Hub An open benchmarking platform for the AI and drug discovery research community to share, upload, exchange, and use research and data; organize conferences and events; and engage in discussion. Available at polarishub.io.
• RxRx3-core A high-quality, standardized dataset for benchmarking and model evaluation in biological perturbation modeling, shared to promote open science and transparent progress in the field.
• Valence Portal A community platform providing the AI and drug discovery research community a space to share research, upload and exchange data, organize events, and engage in discussion.

Quantifiable outcome

• TxPert achieves performance approaching experimental reproducibility levels for predicting unseen single perturbations in multiple cell lines (K562, Jurkat, HEPG2)
• +2 more outcomes

Segments1 record

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability7 records

Feature5 records

Partnerships

Three partnerships are on record, tiered minor and core.

• TempusminorStrategic or Co-development PartnerRecursion announced a data collaboration deal with Tempus to access patient-level data, expanding the biological and clinical data foundation for AI model development. This partnership supports the data scale advantages that enable Valence Labs' virtual cell research.
• NVIDIAcoreTechnology or IntegrationNVIDIA partnered with Recursion to power BioHive-2, described as the most powerful supercomputer in biopharma. This infrastructure partnership provides Valence Labs with the computational capacity essential for training industry-leading AI models at scale.
• University of British ColumbiaminorStrategic or Co-development PartnerAcademic collaboration with UBC Computer Science department on perturbation prediction research. One author of the TxPert paper is affiliated with UBC.

Scale indicators5 records

Recent moves6 records

Expansion highlights5 records

Direct peers

• Atomwise: Atomwise applies AI (AtomNet) to small-molecule drug discovery. It overlaps with Valence Labs' virtual cell and molecular modeling ambitions, particularly in cheminformatics and structure-based prediction.
• Generate Biomedicines: Generate Biomedicines is pioneering generative AI to design novel protein therapeutics. It shares the AI-driven biology thesis with Valence Labs and competes for pharma partnerships and talent in the computational drug discovery space.
• Owkin: Owkin applies AI to patient-level biomedical data and multimodal models for drug discovery and clinical research. It overlaps with Valence Labs in combining foundation models with biological and clinical data, though focused on real-world patient data rather than interventional screens.
• Insitro: Insitro is a machine-learning-driven drug discovery company combining high-throughput biology with ML models for target discovery and patient stratification. It is one of the closest direct competitors to Valence Labs/Recursion in using AI to model cellular responses to perturbations.
• Exscientia: Exscientia is an AI-driven pharma company using ML to design and prioritize drug candidates. It is a direct competitor in AI-led drug discovery with multiple pharma partnerships and clinical-stage assets.
• Isomorphic Labs: Isomorphic Labs (Alphabet/DeepMind spinout) builds AI models for drug discovery, leveraging DeepMind's foundation models (e.g., AlphaFold). It pursues similar virtual-cell-style ambitions with deep compute resources, making it a leading direct competitor in AI-driven biology.
• BenevolentAI: BenevolentAI builds AI models for target identification and drug discovery across multiple therapeutic areas. It is a direct competitor in the AI-driven drug discovery market, with publicly listed status and ongoing pharma collaborations.

Broad incumbents

• Recursion Pharmaceuticals: Recursion is the publicly traded parent of Valence Labs and operates the automated biology/chemistry labs, BioHive supercomputer, and interventional dataset (>65 PB) that Valence Labs builds on. Recursion's pipeline and platform define the strategic mandate for Valence Labs' research.
• Ginkgo Bioworks: Ginkgo Bioworks operates a large-scale automated biology platform for cell engineering and synthetic biology. It is comparable as an industry incumbent with massive wet-lab data generation, though focused on cell engineering rather than drug discovery per se.

Emerging players

• Lantern Pharma: Lantern Pharma applies AI/ML to oncology drug development, using RADR platform for biomarker and drug response prediction. It is an emerging player with narrower therapeutic focus but comparable AI-driven biology approach.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks6 records

Key highlights6 records

Customer concentration

• Xx.com/valence_ai
• GitHubgithub.com/valence-labs
• LinkedInlinkedin.com/company/valenceailabs

Revenue estimate

Valuation estimate

Number of profiles

Profiles1 record

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Valence Labs