San Francisco-based Valinor Discovery has emerged from stealth mode with an ambitious mission to transform drug development through AI-powered patient simulations. The company is building generative machine learning models that predict how patients will respond to therapies before any clinical testing begins, potentially addressing the high failure rates that plague pharmaceutical development.
"Clinical drug development is failing us," said Joshua Pacini, Co-Founder and CEO of Valinor. "Despite consuming more than 70% of pharma R&D budgets, 90% of drugs still fail in clinical trials. We believe there's a better way."
Bridging the Gap Between Lab and Clinic
At the core of Valinor's approach is the creation of virtual patient models trained on matched multi-omics and clinical assay data from individual patients. The company plans to release its first oncology-focused model within six months, which Pacini describes as "the first to accurately simulate the holistic physiological impact of a chemical on cancer progression."
This represents a significant shift from virtual cell models to virtual patient models, with simulations designed to predict both molecular-level changes (transcriptomic, protein abundance, methylation) and clinical assay outcomes. These capabilities enable compound prioritization, biomarker identification, and patient stratification before entering expensive clinical trials.
"By linking molecular perturbations to actual clinical assay data, drug hunters can test therapies virtually in a matter of weeks before investing years at the bench or in the clinic," explained Zhanel Nugmanova, Co-Founder and Chief Scientific Officer at Valinor.
Proprietary Data Generation
Valinor is addressing what Pacini identifies as a critical industry challenge: "The scarcity of high-quality clinical data significantly limits the development of translatable biological models." To overcome this barrier, the company is generating proprietary longitudinal datasets that integrate transcriptomic, proteomic, and methylation profiles with clinical assay results and primary cell measurements from the same patients.
These matched datasets—from cellular samples to biopsies—are explicitly designed to enable patient-level simulation of drug performance in real-world scenarios, not just in vitro conditions.
Strategic Collaborations
To validate and extend its platform, Valinor has established key academic and industry partnerships:
Valinor is collaborating with Professor Fabian Theis and the Computational Health Center at Helmholtz Munich to develop benchmarking standards for perturbation models. These benchmarks will support evaluation across therapeutic areas and drug types, with plans to share them via the OpenProblems platform.
"Valinor's approach represents an exciting step forward in predictive drug development," said Dr. Theis. "We're excited to work with them to create more robust, clinically translatable benchmarks."
A separate partnership with the Montgomery Lab at Stanford Medicine will focus on applying Valinor's models to evaluate compounds targeting Alzheimer's disease. The lab, led by Professor Stephen Montgomery, has contributed to large-scale transcriptomic studies, including the Genotype-Tissue Expression (GTEx) project.
Platform Access and Integration
Valinor offers its platform through a hosted interface developed with Latch Bio, providing no-code access to generative models and workflows for biopharmaceutical developers. The platform supports multiple use cases, including:
- Hit-to-lead analysis
- Biomarker panel selection
- Patient stratification
- Adverse effect modeling
- Clinical endpoint simulation
- Dose-level modeling
- Disease association analysis
- Drug repurposing
Ready-to-use workflows are available out of the box, and Valinor offers bespoke retraining to align models with a sponsor's asset, indication, or trial protocol.
Industry Focus and Advisory Board
Importantly, Valinor does not develop its own therapeutics. "Everything we build is designed to power our pharmaceutical customers," stated Pacini. The company's goal is to equip biopharma partners with tools to bring more effective drugs to market faster and at lower cost.
To guide its development, Valinor has assembled an advisory board with expertise spanning machine learning, clinical operations, and drug discovery:
- Stephen Montgomery, PhD – Stanford University
- Chase Neumann, PhD – Associate Director of Oncology, Recursion
- Bryan Norman, PhD – Formerly of Eli Lilly and Enveda Biosciences
- Tim Sullivan, PhD – Chief Business Officer, Infinimmune
- Matt Donne, PhD – Formerly of Spring Science
Market Impact and Future Directions
By enabling the simulation of drug efficacy before clinical trials begin, Valinor's platform has the potential to significantly reduce the time and cost of bringing new therapies to market. The company is also collaborating with -omics sequencing companies to develop custom models trained on data from biopharma-led clinical trials, both ongoing and historical.
As pharmaceutical companies continue to seek ways to improve R&D productivity and reduce clinical failure rates, Valinor's approach represents a promising new direction in the application of AI to drug development—one that could help address the industry's persistent challenge of translating laboratory findings into clinical success.
