Unlock Biology

Tahoe Therapeutics is an AI-forward drug discovery biotech building very large single-cell atlases and training models to predict cellular responses. In February 2026, coverage highlighted its high-throughput scRNA-seq atlas generation and multiple parallel cancer programs. In January 2026, Tahoe, Arc Institute, and Biohub announced a landmark initiative to generate a very large perturbation-rich single-cell dataset for virtual cell models. This combination of huge single-cell throughput, perturbations, and multi-program discovery creates strong demand for foundation-model style, transferable predictive modeling.

Position Unlock Biology as complementary “foundation model infrastructure” to turn Tahoe’s growing single-cell/perturbation datasets into reusable, cross-program predictive priors (cell-state transitions, MoA, and response to perturbations). Emphasize scalable multimodal fusion readiness (adding spatial, imaging, CRISPR screens) and model transfer across tumor contexts to reduce experimental cycles and failed targets. Entry angle: pilot on a subset of Tahoe’s atlas + one active oncology program to demonstrate improved experimental prioritization and generalization across perturbations.

Cartography Biosciences is an oncology biotech building an atlas-driven discovery engine to identify tumor-selective antigens and antigen pairs. In January 2026, reporting covered Pfizer’s collaboration with Cartography leveraging its Atlas and Summit platforms to discover and validate tumor-selective antigens. The company’s approach implies heavy single-cell/tissue profiling and complex target validation across heterogeneous tumors, where generalizable cell models can reduce iteration cycles. The presence of a computational biology function is implied by their platform and hiring for computational biology roles.

Approach with a message about improving transfer learning and cross-cohort generalization: model tumor/immune cell-state responses and antigen co-expression patterns to de-risk target selection earlier. Entry angle: deploy Unlock Biology’s foundation model as a standardized representation layer across Cartography’s Atlas datasets to predict which antigens stay tumor-selective across patient subtypes and microenvironments. Offer a secure, fine-tunable model workflow to integrate new sponsor/partner datasets while preserving proprietary data boundaries.

PharosAI is a UK research consortium aiming to build a very large multimodal cancer dataset and pair it with AI models and analytical capabilities. On February 5, 2026, 10x Genomics announced PharosAI will use Xenium spatial to transform archived NHS cancer samples into high-resolution multimodal datasets for AI models to support diagnosis and drug discovery. This implies immediate needs in multimodal integration, scalable model training, and foundation-model style representations across cohorts and tissue contexts. As an initiative spanning multiple institutions, standardization and reusable model backbones are critical to avoid fragmented analyses.

Approach as an infrastructure partner: Unlock Biology can provide a generalizable cell-state foundation model layer that PharosAI can fine-tune across cancer types, tissue sites, and assay versions to keep outputs consistent. Entry angle: propose a reference model for tumor microenvironment states integrating spatial transcriptomics with other modalities, enabling transfer to new cohorts and rapid hypothesis generation. Emphasize governance-aware deployment (federated/secure compute options) to fit NHS data constraints.

Immunai is a biotech focused on comprehensively mapping and decoding the immune system with single-cell biology and AI to power therapeutic discovery and accelerate drug development. In January 2026, coverage described Immunai expanding its single-cell multi-omics efforts to grow its AI atlas, reflecting ongoing data volume growth and integration needs. Their platform orientation (AI + single-cell) aligns closely with the requirement for transferable representations and predictive modeling across cohorts and indications. This is a strong “platform buyer” profile that values reusable models over one-off analytics.

Position Unlock Biology as a complementary foundation-model approach that can expand modality coverage (e.g., spatial + imaging + perturbations) and improve transfer across new therapeutic areas. Entry angle: benchmark Unlock’s cell foundation representations against Immunai’s internal embeddings on a shared task (e.g., response prediction / state change under perturbations), then propose a fine-tuning pipeline for partner datasets. Emphasize generalization across assays and the ability to standardize across multiple sponsor programs.

Omics Empower is a global multi-omics service provider with explicit single-cell and spatial transcriptomics capabilities. On February 13, 2026, it announced Xenium in situ spatial transcriptomics services at its Hong Kong Science Park lab and described support for Visium and Stereo-seq as well. Its LinkedIn presence confirms focus on single-cell and spatial services with a large organization footprint. As a CRO serving many sponsors, it likely faces repeated integration demands across modalities and would benefit from a standardized predictive modeling layer to shorten analysis cycles and improve experimental prioritization for clients.

Position Unlock Biology as a reusable, sponsor-agnostic modeling backbone that Omics Empower can apply across client projects (predict response/MoA and prioritize perturbations) rather than bespoke analysis each time. Entry angle: co-develop a “predictive multiomics package” for oncology/immunology clients combining Xenium/Visium + scRNA-seq and deliver model-derived hypotheses (cell-state transitions, microenvironment drivers). Emphasize secure deployment and fine-tuning per sponsor dataset while keeping Omics Empower’s service workflow streamlined.

Sirona Dx is a technical CRO offering high-complexity single-cell and multi-omics services, explicitly including spatial biology. Its CosMx service page describes RNA + protein detection with subcellular resolution, indicating multimodal integration needs. The LinkedIn company page states it provides advanced single cell, multi omics services and supports biopharma regulatory requirements (CLIA/GCLP positioning). As a CRO with multiple sponsors, a reusable predictive modeling layer could reduce cycle time and make outputs more operational for target validation and biomarker programs.

Pitch Unlock Biology as the “predictive layer” that sits on top of Sirona’s spatial multiomics services—turning RNA+protein spatial readouts into mechanistic predictions and prioritization (e.g., resistance drivers, responder stratification). Entry angle: joint offering for immuno-oncology trials—model tumor-immune spatial states and predict response/toxicity signatures across cohorts. Emphasize portability across sponsors while respecting data governance through fine-tuned, per-client models.

Via Scientific provides Foundry, an enterprise-scale multi-omics data processing and analytics platform with explicit single-cell analytics and spatial biology support. Its LinkedIn posts cite expansion to pipelines/tools like Visium, Visium HD, and Xenium, connecting single-cell and spatial in one environment. This indicates active multimodal integration demand and a computational-biology-forward operating model. Foundry’s platform focus suggests readiness to operationalize model outputs, making it a natural distribution path for foundation-model predictions embedded into pipelines.

Partner rather than sell point-by-point: propose integrating Unlock Biology’s foundation-model inference as an optional module inside Foundry (API-first), enabling users to run predictive tasks (state transition, perturbation response, MoA clustering) as part of standard workflows. Entry angle: co-market a ‘predictive multiomics’ pipeline for oncology/immunology where outputs are standardized embeddings and prioritization scores. Highlight that Unlock’s models can generalize across datasets, reducing bespoke model engineering per customer.

Tempus is an AI and precision medicine company offering integrated multi-omics capabilities (including single-cell and spatial methods listed in its omics solutions). On March 4, 2026, Tempus announced an expanded multi-year collaboration with Merck to accelerate biomarker discovery and development supporting Merck’s oncology portfolio. This indicates active, large-scale multimodal dataset work and pressure to deliver predictive biomarkers and stratification—aligned with predictive modeling layers rather than descriptive reporting. Tempus also publicly describes access to single-cell and spatial platforms (e.g., 10x Visium and Chromium) as part of its life sciences omics offerings.

Position Unlock Biology as a specialized foundation-model layer for cellular-state prediction that can plug into Tempus’s multimodal biomarker pipelines, especially where single-cell/spatial data are used to explain heterogeneous response. Entry angle: propose a joint pilot on a translational oncology dataset to generate transferable cellular embeddings for patient stratification and MoA inference, with strong auditability for regulated environments. Emphasize improved generalization across cohorts and modalities compared with study-specific models.

Vizgen develops in situ single-cell spatial genomics (MERFISH/MERSCOPE) and protein assays, with AI-driven image analysis (STARVUE). On February 19, 2026, Vizgen announced a spatial multi-omics roadmap at AGBT 2026, including collaborations around spatial molecular profiling of organoids—implying expansion in modality mix and translational model systems. Their technology ecosystem creates downstream pressure for integrated, predictive modeling across spatial transcriptomics + protein + organoid readouts. While a tools vendor, Vizgen is increasingly tied to translational and drug discovery use cases where model generalization across assays is valuable.

Position Unlock Biology as a partner-grade foundation modeling layer that makes Vizgen-generated spatial multiomics data more predictive (cell-state transition, response prediction, MoA inference) and easier to operationalize in biopharma collaborations. Entry angle: co-develop reference models for organoid spatial profiling and demonstrate cross-assay transfer (MERFISH + protein + imaging). This can help Vizgen differentiate beyond instrumentation by enabling AI-ready predictive biology workflows for its biopharma customers.

Illumina is expanding into data/software/AI with its BioInsight division and introduced the Billion Cell Atlas, a genome-wide perturbation dataset built to accelerate AI-driven drug discovery. The January 13, 2026 press release states the Atlas is being built under an alliance framework with AstraZeneca, Merck, and Eli Lilly as founding participants. This is a direct signal of step-change in single-cell data volume plus perturbation context, and a strategic emphasis on predictive models rather than analysis of isolated datasets. While Illumina is a tools/data vendor, its BioInsight direction suggests active partnering needs for foundation-model scale biology and predictive use cases across modalities.

Approach as a strategic partner rather than customer: propose embedding or validating Unlock Biology’s foundation-model approaches against Billion Cell Atlas perturbation tasks (predicting cell-state shifts, MoA signatures, and generalization across cell lines). Entry angle: demonstrate cross-modality extensions (e.g., adding spatial priors or imaging alignment) and provide model evaluation frameworks that make outputs more actionable for pharma participants. This can strengthen Illumina’s BioInsight narrative and increase adoption by pharma teams seeking reusable predictive layers.