Prioritizing Candidate Genetic Variants Using GTEx, HuBMAP, and IDG

A reproducible CFDE training module

Use tissue, cell-type, and protein data to select candidate genetic variants for further study.

Estimated total module time: Approximately 2 hours 45 minutes

Start with the published candidate genetic variants, then examine tissue expression, cell-type expression, and protein information to choose variants for follow-up.

If a term is unfamiliar, refer to the Glossary.

NoteBefore you begin

This module assumes basic familiarity with genes, genetic variants, and tabular data. The lessons introduce the API concepts used in each activity. You should be comfortable running a code cell and recognizing rows, columns, filters, and sorted results.

These activities run in a web browser and request live data, so results may change over time.

Dated teaching files support the activities during a service interruption. You can also run the matching notebooks in Jupyter.

The research question

Which published candidate genetic variants should be prioritized for further study, and which experimental models could be used to test them?

This module uses data from a 2025 publication that reports results from a whole-genome sequencing study of early-onset advanced heart failure (PMID: 39910139). The researchers studied 102 Swedish heart transplant or heart-assist-device recipients.

They used a computational pipeline to rank variants and clinical experts to classify them. Their Supplementary Table S4 reports 54 variant rows from 46 people across 25 genes.

Variant analysis can produce a large list of candidate genetic variants. Even after evaluation by a clinical geneticist or variant analyst, a research group may still need to decide which findings to investigate first and how to test them.

If you want to learn more about variant identification and analysis at an introductory level, see EMBL-EBI’s Variant identification and analysis lesson.

The prioritization lesson uses ProtVar, an EMBL-EBI resource for human missense variants, to compare complementary predictions for TNNT2 p.Asp259Ala. AlphaMissense provides a predicted-pathogenicity score, EVE evaluates the substitution against evolutionary patterns in related protein sequences, and FoldX estimates its effect on protein stability.

The module uses six evidence questions to guide prioritization:

Evidence dimension Question
Published evidence What did the source study report about the exact variant and phenotype?
Tissue expression Is the associated gene expressed in a relevant heart tissue?
Cell-type expression Is the gene detected in an appropriate cardiac cell type?
Protein knowledge What biological, disease, or therapeutic information is available for the protein?
Variant-specific evidence What is known or predicted about the exact substitution?
Follow-up feasibility What experiment or experimental model could test the hypothesis?

Together, these evidence types support a qualitative prioritization decision.

Tutorial workflow

flowchart LR
    %% Begin with the variants already evaluated in the source paper.
    variants["Published variants<br/>54 rows, 25 genes"]

    %% Keep the three evidence sources visually distinct.
    gtex["GTEx<br/>Tissue expression"]
    hubmap["HuBMAP<br/>Cell-type expression"]
    pharos["Pharos / IDG<br/>Protein annotations"]

    %% Use the combined results to choose focused research follow-up.
    shortlist["Shortlist of candidate genetic variants<br/>for follow-up"]

    %% Query each resource independently, then bring the results together.
    variants --> gtex
    variants --> hubmap
    variants --> pharos
    gtex --> shortlist
    hubmap --> shortlist
    pharos --> shortlist

    classDef source fill:#f6f7f7,stroke:#63666a,color:#141415
    classDef evidence fill:#cce4ef,stroke:#3d64b3,color:#141415
    classDef outcome fill:#764c82,stroke:#5e3d68,color:#ffffff

    class variants source
    class gtex,hubmap,pharos evidence
    class shortlist outcome

What you will learn

By the end of this module, you will be able to:

  1. Explain what an application programming interface, or API is and how APIs support repeatable research.
  2. Use GTEx to check whether genes linked to candidate genetic variants are expressed in a disease-related tissue.
  3. Use HuBMAP to study gene expression in specific cell types while accounting for missing data and small groups.
  4. Use Pharos to interpret protein information and IDG Target Development Levels.
  5. Combine several data sources to prioritize candidate genetic variants for follow-up.

Data sources

Resource What it is How this module uses it
Candidate genetic variant table and source paper The local table contains all 54 rows from the paper’s Supplementary Table S4. Provides the published HGVS names, study scores, classifications, phenotypes, and comments.
GTEx v10 GTEx is an NIH Common Fund reference resource for studying gene expression and genetic regulation across human tissues. A GTEx Portal REST API response provides median expression in heart atrial appendage and left ventricle.
HuBMAP HuBMAP is an NIH Common Fund program that maps cells and molecules within human tissues. A HuBMAP Cells API response provides expression summaries for selected heart cell types.
IDG and Pharos IDG is an NIH Common Fund program focused on understudied druggable proteins. Pharos is its integrated protein-target information resource. A Pharos GraphQL API response provides protein information and Target Development Levels.
Integrated teaching table A local table created by joining the published variants with the saved GTEx, HuBMAP, and Pharos data by gene symbol. Adds GTEx, HuBMAP, and Pharos information to all 54 variant rows while keeping the source of each result clear.

Why combine these resources?

Combining these resources helps us move from a list of candidate variants to a testable follow-up plan. The combined results help us choose a candidate, an experimental model, and an assay that addresses the proposed biological effect.

Rosalution provides another example of combining annotations from several sources to prioritize candidate genetic variants. It keeps each annotation linked to its source.

How to use the lessons

Work through the pages in order.

  • Activities: Run the same analysis code in the website or in the matching notebook.
  • Knowledge checks: Select an answer and review the feedback.
  • Complete workflow: Use the integrated notebook to run the complete analysis in one place.

Next: Learn how APIs provide repeatable access to biomedical data.

View Jupyter notebook