Calypr: Computation Analysis and Learning Yielding Predictions and Reports

Challenge: Data Silos + Duplications + Interoperability

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

. . .

lab A

lab Z

Data Lake

UI interface

Search API UI

Clark et al. AI-readiness for Biomedical Data Bridge2AI Recommendations

Management

ETL

Standards

Data Science Garage: Data Standards + ETL + management

Calypr : Extending Gen3 with Microservices for Bioinformatics Data Storage, Querying & Management

Modular microservises

Gen3 User: Build cohort and download
Gen3 Developer: build + integrate micorservises
Gen3 operator: Spinup an instance + configure data + costumize frontend

BMEG Iceberg + FHIR graph schema

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Gene

Allele

Protein

Transcript

Pathway

Calypr: Data Ingestion + Integration

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Calypr: Landing and Main App Page

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Data Explorer
Storage Manager
Custom Report Pages
Data Viewer
Interactive Computation
Graph Queries
Project Management

Calypr: Discovery and Summary Reports page

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Calypr: File Viewer App

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Calypr: Data Explorer - Cohort Builder

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Calypr: Data Explorer - Temporal Data Tracking

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Calypr: Image Viewer

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Calypr: Connecting Applications

Data can become fragmented locked in silos, duplicated unnecessarily, and difficult to integrate.

Nasim Sanati
Computational Biologist
@teslajoy

Quinn Wai Wong
Research Software Eng.
@quinnwai

Kyle Ellrott
PI + Associate Professor
@kellrott

Liam Beckman
Research Software Eng.
@lbeckman314

Brian Walsh
Sr. Research Software Eng.
@bwalsh

Jordan Tagle
Computational Biologist
@jordan2lee

Matthew Peterkort
Research Software Eng.
@matthewpeterkort

Data Science Garage Team + Creason's Team

Allison Creason
Sr. Computational Biologist

Jay Egger
Computational Biologist

Code + Stunning Sunrise + Local Coffee + Great Team

Extra Slides

HTAN: Bidirectional Transformer between FHIR and HTAN schema.

GRIP & FHIR

Querying TCGA BRCA Data with subtype labels for classification

Subtype Classification with Neural Networks

ML modeling and Shapley Feature Extraction

Pathway Enrichment Analysis

Extracting Biological Insights from Gene Set Results

Where We Work

Knight Cancer Research Building @ South Waterfront

Across from OMSI and right by Tilikum Crossing

Scientific Problem

How can we find actionable insights in a disease as complex as cancer?

Scientific Problem

How do we integrate data to extract meaningful insights?

ETL (Extract, Transform, Load)
- FHIRizer – Transforms GDC, Cellosaurus, HTAN data into FHIR graph representation.
- Snakemake & Airflow – Automate and orchestrate ETL workflows.

Bringing It All Together with Python

Data Standards, ETL, and Management

Management
- GRaph Integration Platform (GRIP) - a powerful graph query engine designed to span across multiple database backends.
- Gen3 – Data commons for storage, querying, and management.

Standards
- Pydantic – Validates and enforces FHIR schema.
- GA4GH (Global Alliance for Genomics & Health) – Establishes global standards for genomic data interoperability.
- Pathway Databases (Pathway Commons, Reactome) – Structure & standardize biological context through molecular interaction graphs.