Omica Cancer Atlas

Multi-omic tumor cohorts from Latin America.

Tissue specimens matched with digital pathology, multi-omic sequencing, and longitudinal patient outcomes.

Request a cohort Explore specimen characterization
  1. 01 Histopathology
  2. 02 Genomics
  3. 03 Transcriptomics
  4. 04 Proteomics
  5. 05 Clinical
  6. 06 Population

Specimen characterization

One tissue block. Complete multimodal data.

All modalities originate from the same biopsy block and verified clinical records, eliminating linkage errors.

  1. 01

    Histopathology

    40x whole-slide imaging across brightfield H&E and IHC. Comprehensive nuclear segmentation with 630 extracted morphologic features per cell.

    SVS · NDPI · DICOM

  2. 02

    Genomics

    Whole-exome sequencing and targeted oncologic panels with somatic mutation and copy-number variant calling.

    FASTQ · VCF

  3. 03

    Transcriptomics

    Bulk RNA sequencing and spatially resolved transcriptomics mapped to slide coordinates.

    Counts matrices · spatial files

  4. 04

    Proteomics

    Multiplex immunofluorescence and spatial proteomics with quantified per-cell protein expression.

    Marker intensities per cell

  5. 05

    Clinical

    Comprehensive longitudinal records: treatment regimens, therapeutic response, recurrence intervals, and overall survival.

    De-identified per standardized regulatory protocols

  6. 06

    Population

    Median 65% Indigenous American ancestry, addressing critical representation gaps in TCGA and international genomic references.

    Continuously expanding clinical pathology network

Access models

Train on existing cohorts, validate on held-out data, or collect prospectively.

Define your cohort requirements by tumor type, required omic layers, ancestry thresholds, and target clinical endpoints.

  • 01

    Training cohorts

    Analysis-ready multi-omic cohorts formatted for machine learning, available through exclusive or non-exclusive licensing.

    Exclusive or non-exclusive · federated access

  • 02

    Validate and benchmark

    Evaluate model generalizability on sequestered evaluation sets that remain unexposed to public training pipelines.

    How held-out benchmarking works

  • 03

    Prospective cohorts

    Commission targeted cohorts specifying histologic subtype, prior therapy exposure, or longitudinal outcome criteria across active network sites.

    Custom prospective collection

Governance and compliance

Consented, de-identified, and licensed for commercial discovery.

Consent
Broad informed consent for biomedical research under institutional review board oversight.
De-identification
Clinical and demographic records de-identified in strict adherence to regulatory privacy frameworks.
Licenses
Commercial and academic licensing structures tailored to research use or algorithmic IP generation.
Federated access
In situ federated compute environments for secure model training and evaluation without raw data transfer.
Held-out sets
Sequestered benchmark cohorts maintained under strict access controls to prevent benchmark contamination.

Clinical integration

An atlas fueled by daily hospital pathology.

Network laboratories process routine surgical resections and biopsies within PRISMA. Consented cases undergo multi-omic characterization and enrich the Atlas, providing ongoing cohort growth and prospective collection capability.

PRISMA, the pathology workspace Diagnostic programs

Data licensing

Frequently asked questions

What formats are delivered?

Digital whole-slide images are provided as standard SVS, NDPI, or DICOM files with single-cell segmentations and morphological feature matrices. Genomic data is delivered as raw FASTQ and annotated VCF files. Transcriptomics includes gene count matrices and spatial coordinate mappings. Proteomics includes normalized single-cell protein expression values. Clinical histories are fully de-identified in accordance with our regulatory framework protocols. Where data transfer is restricted, secure federated compute environments are available.

What is the typical delivery timeline for a cohort?

In-stock cohorts matching existing Atlas specimens can be packaged and transferred within 2 to 4 weeks following data licensing execution. Prospective collections follow milestones scoped during cohort design.

Are cohorts licensed for commercial AI model development?

Yes. Commercial model training licenses grant broad operational rights to develop, validate, and deploy proprietary algorithms and diagnostic tools.

Can our team secure exclusive cohort access?

Yes. Exclusive licensing arrangements ensure designated cohorts or specific sub-populations remain restricted to your organization for defined therapeutic indications and timeframes.

What options exist if institutional policy prohibits raw data transfer?

Omica provides containerized federated learning environments. Your engineering team deploys training or evaluation code directly to our secured local compute infrastructure, exporting only trained weights and validation metrics.

How does held-out benchmarking operate?

Sequestered benchmark cohorts are held in strict blind storage, completely unexposed to public domain references or training pipelines. Your model container executes against the held-out dataset, generating objective, contamination-free performance audits.

Can Omica prospectively collect rare or customized oncology cohorts?

Yes. Our active clinical pathology network enables prospective screening and acquisition of targeted cohorts according to granular histopathologic subtypes, molecular biomarker profiles, or specific therapy responses.

How are cohort licenses structured and priced?

Pricing is determined by specimen volume, requested analytical modalities, licensing exclusivity, and prospective acquisition requirements. Contact our scientific partnerships team to review your cohort specification.