Valinor Virtual Patient Platform

The Valinor Virtual Patient Platform is a machine learning-based system designed to simulate therapeutic efficacy at the whole-patient level, with the goal of reducing the high failure rate of clinical drug development. The platform is aimed at pharmaceutical and clinical research organizations seeking to predict how therapies will perform before human trials begin.

Valinor's approach centers on building virtual patient profiles from large-scale, longitudinal clinical datasets rather than modeling isolated cell behavior. The platform integrates a broad range of multi-omics and clinical data types to construct these profiles, positioning patient-level simulation as a more predictive alternative to single-cell or siloed experimental readouts.

Supported Data Modalities

• PBMC single-cell RNA sequencing (scRNA-seq)
• Tumor single-cell RNA sequencing (scRNA-seq)
• PBMC single-cell ATAC sequencing (scATAC-seq)
• Tumor single-cell ATAC sequencing (scATAC-seq)
• Whole genome sequencing
• Blood plasma proteomics
• Spatial omics
• Radiology imaging
• Histopathology
• Cell viability assays
• Cell imaging data
• Circulating tumor DNA (ctDNA)
• Treatment history
• Structured clinical data

Core Platform Capabilities

• Generates virtual patient profiles grounded in rich, longitudinal multi-omics and clinical assay data
• Simulates therapeutic efficacy at the whole-patient level rather than at the level of isolated cell experiments
• Runs automated patient simulations trained on large-scale longitudinal clinical datasets
• Models how therapies are expected to perform prior to the initiation of human clinical trials

Context and Motivation

• Clinical trials represent approximately 70% of pharmaceutical R&D costs
• Approximately 90% of clinical trials fail, making them the primary bottleneck in drug development
• The platform is designed to address this failure rate by providing pre-trial simulation and prediction capabilities

Valinor describes its models as being trained on what it characterizes as the largest clinical longitudinal datasets assembled for this purpose. The platform is developed in collaboration with external research partners and applies machine learning methods to integrate the multi-modal data types listed above into unified virtual patient representations.