Stream 2: Genomics Signal Integration

Wed, 01 Jan 2025 00:00:00 +0000

Overview

Stream 2 is the data backbone of the QuantOmics pipeline. Trainees in this stream tackle the grand challenge of converting noisy, high-dimensional quantum sensor data into high-fidelity, analytically tractable multi-omic datasets. Without Stream 2, the hardware signals from Stream 1 cannot be interpreted, and the AI models of Stream 3 have no validated data to learn from.

The stream bridges two worlds: the messy, real-world signals from quantum probes operating in complex biological matrices, and the clean, structured representations required by state-of-the-art computational genomics and AI methods.

Research Focus Areas

Whole-Genome Sequencing & Variant Analysis

Long- and short-read sequencing data generated by quantum-enhanced nanopore platforms requires sophisticated analysis. Trainees work on:

Applying comprehensive whole-genome sequencing (WGS) analyses to brain organoids to identify therapeutic targets for neurodevelopmental disorders such as autism
Developing improved variant calling algorithms that account for unique noise characteristics of quantum-coupled sequencing
Building tools for structural variant detection in complex genomic regions

Multi-Omic Data Fusion

Integrating data across molecular scales — from DNA to protein to cellular phenotype. Key projects include:

Fusing electrical and optical sensor data streams with nanopore sequencing reads and methylation maps
Building end-to-end Snakemake/Nextflow pipelines for reproducible multi-omic analysis
Developing early vs. late fusion strategies for genomic, transcriptomic, proteomic, and EHR data
Applying representation learning to extract shared latent features across data modalities

Computational Tools for Sensor-Derived Genomic Data

Novel sensors generate novel data formats. Stream 2 trainees develop new bioinformatics tools:

Signal processing algorithms for denoising raw quantum sensor output before genomic analysis
Probabilistic models for base-calling from quantum-coupled nanopore reads
Quality control frameworks tailored to attomolar-sensitivity assay data

Epigenomics & Methylation Analysis

DNA methylation patterns hold rich information about cell state and disease. Trainees build:

Pipelines for integrating methylation data with quantum sensor readout from epigenetic biosensors
Differential methylation analysis in disease-relevant organoid models
Tools for cross-referencing methylation signatures with EHR phenotype data

Validation Against Real Disease Models

Stream 2 research is validated against clinically relevant biological systems, primarily using patient-derived organoids developed in collaboration with Stream 1. Key application areas:

Neurodevelopmental disorders — applying WGS to brain organoids to identify new therapeutic targets for autism, building on existing team expertise in the genomic architecture of these conditions
Cancer — integrating multi-omic data to characterize tumor heterogeneity and identify drug-resistant cell populations in organoid cultures
Cardiotoxicity — developing computational pipelines for analyzing multi-omic data from cardiomyocyte organ-on-chip models under drug perturbation

Example Trainee Projects

Whole-genome sequencing of autism brain organoids — identifying novel de novo variants and gene regulatory disruptions associated with autism spectrum disorder
Snakemake pipeline for multi-omic fusion — building a reproducible, containerized pipeline integrating nanopore sequencing, ATAC-seq, and EHR metadata
Methylation-guided biomarker discovery — developing an algorithm that identifies disease-specific methylation patterns detectable by quantum epigenetic biosensors
Cross-ancestry variant classification — building genomic models that integrate population-diverse genetic marker data to reduce bias in variant pathogenicity prediction

Equity & Diversity in Genomic Data

A critical challenge in Stream 2 is the profound lack of diversity in genomic reference databases, which are overwhelmingly of European origin. Trainees in this stream are explicitly trained to:

Integrate data representing comprehensive genetic markers from diverse populations
Understand how racial bias in genomic AI can lead to misclassification for underrepresented groups
Apply ethical frameworks for Indigenous data sovereignty and community consent in genomic research
Validate tools across diverse cell lines to ensure equitable performance

Stream 2 Co-Leads

Dr. Brett Trost (UofT / SickKids) — Computational genomics, multi-omic pipeline development
Dr. Jacques Corbeil (U Laval / MILA) — Medical genomics, AI-driven data integration, organ-on-chip resources
Dr. Brenda Andrews (UofT) — Functional genomics, systems biology, gene networks (Tier 1 CRC)

Courses Supporting Stream 2

Course 1.3 — AI in Genomics
Bootcamp 1.4 — Multimodal-Omics Data Integration (the core course for this stream)
Course 1.5 — Responsible Innovation & EDI in Precision Health (Indigenous data sovereignty, bias in genomic AI)

Computational Biology | QuantOmics