Medical Genomics: Genomic Repositories and Immune Repertoire Mining

Page 1: Course Overview and Presenter Information

• Medical genomics presentation delivered on September 30, 2026, by Jesper Madsen, PhD, with materials courtesy of George Blanck, PhD (USF Health).

Page 2: Learning Objectives

• Understand the organizational structure of the NIH database of Genotypes and Phenotypes (dbGaP).
• Compare and contrast the Genomic Data Commons (GDC) with cBioPortal.
• Identify major international repositories, including European databases and the Chinese Glioma Genome Atlas.
• Review an applied medical genomics investigation utilizing raw sequencing files.

Page 3: Reference Documentation

• Recommended reference: Genomic Data Commons (GDC) User's Guide on Projects (https://docs.gdc.cancer.gov/Data_Portal/Users_Guide/Projects/).

Page 4: Foundational Genomic Definitions

• Exome: Complete sequence of all $235,000$ coding exons.
• Genome: Total chromosomal DNA sequence consisting of $3 \times 10^9\,\text{base pairs}$.
• RNASeq: mRNA expression profiling across all genes for a specified sample.
• Additional Profiling Modalities: Cytosine methylome, RNA microarrays, mass spectrometry proteomics, and copy number variation (CNV) via single nucleotide polymorphism (SNP) arrays.

Page 5: European Genomic Repositories

• COSMIC: Catalogue of Somatic Mutations in Cancer.
• European Nucleotide Archive (ENA): Non-controlled biological data, including animal sequencing.
• European Genome-phenome Archive (EGA): Controlled-access repository for human patient files.
• UK Biobank: Large-scale prospective health and genetic resource.

Page 6: Chinese Glioma Genome Atlas (CGGA)

• CGGA: Specialized clinical and genomic repository focused on molecular profiling and data dissemination for gliomas in Chinese cohorts.

Page 7: NCBI Sequence Read Archive (SRA)

• Centralized NCBI repository hosting raw sequencing read data from platforms such as Illumina NextSeq 500 and HiSeq 4000 across diverse taxa (Homo sapiens, Mus musculus, Canis lupus familiaris).

Page 8: Database of Genotypes and Phenotypes (dbGaP) Overview

• NIH umbrella organization archiving studies investigating genotype-phenotype interactions in humans.
• Provides a public catalog alongside controlled-access phenotypic and genetic datasets requiring NIH authorization.
• Archives large cohort projects, including The Cancer Genome Atlas (TCGA).

Page 9: dbGaP Portal Architecture

• Maintained by the National Center for Biotechnology Information (NCBI) and National Library of Medicine (NLM) (www.ncbi.nlm.nih.gov/gap) to archive and distribute genotype-phenotype association data.

Page 10: dbGaP Search Query: Parkinson's Disease Cohorts

• Parkinson's disease queries in dbGaP yield $28$ studies.
• Representative entry: phs001172.v1.p2 (NINDS Parkinson's Disease), a case-control whole exome sequencing study in North American Caucasian cohorts.

Page 11: NINDS Parkinson's Disease Substudies

• Umbrella study phs001172.v1.p2 contains distinct sub-studies:
  • phs000089.v4.p2: Genome-Wide Genotyping in Parkinson's Disease.
  • phs001103.v1.p2: Exome Sequencing in Parkinson's Disease.

Page 12: Exome and Genotype Data Quantification in dbGaP

• Substudy phs001103.v1.p2 dataset distribution under General Research Use (GRU) consent:
  • SNP/CNV genotypes via next-generation sequencing: $618$ subjects and $618$ samples.
  • Whole exome sequencing (WXS): $618$ subjects and $1,223$ sample files.

Page 13: Repositories Overview: GDC vs. cBioPortal

• Genomic Data Commons (GDC): Controlled-access repository distributing raw sequencing reads.
• cBioPortal (cbioportal.org): Publicly accessible repository providing curated, processed genomic data.

Page 14: Data Processing Hierarchy: Raw vs. Processed Datasets

• Raw Data: Primary unaligned or aligned instrument reads (DNA or RNA) from tumor specimens and matched normal tissues; distributed through GDC.
• Processed Data: Curated molecular profiles and clinical attributes confined to tumor specimens (excluding normal tissue controls); distributed through cBioPortal.

Page 15: Categories of Curated Genomic Data

• Somatic mutations derived from tumor exomes.
• Transcriptome profiles generated by RNASeq.
• Cytosine methylation patterns across the genome.
• Copy number variations (such as $MYCN$ amplification in neuroblastoma and $DHFR$).
• MicroRNA expression levels.
• Clinical phenotypic and outcome data.

Page 16: cBioPortal Query Workflow

• Selection of disease cohorts (e.g., Colorectal Adenocarcinoma, Dana-Farber Cancer Institute [DFCI], Cell Reports 2016; $619$ patients/samples).
• Direct query submission of gene lists using official HUGO symbols to inspect mutation landscapes.

Page 17: Tumor Suppressor Panel and Microsatellite Instability (MSI)

• Evaluated tumor suppressor panel ($48$ genes): AKAP12, AXIN1, BMP2, BMPR1B, BMPR2, BRCA1, BRCA2, BRMS1, BRMS1L, CASZ1, CDKN2A, CHD5, WRB, CHD6, CHEK2, CTCF, DLC1, DLEC1, DYNLL1, DOK2, FLCN, FOXP3, GPR68, ING1, ING4, INPP4B, KISS1, KISS1R, KLF6, LATS2, LIMD1, MAP2K4, MED23, PBRM1, PEBP1, PRDM2, PTEN, RB1, RECK, SMAD4, SMAD7, SMARCB1, SP100, TFPI2, TMPRSS11A, TXNIP, VHL, and WWOX.
• Microsatellite Instability (MSI): Hypermutable state resulting from inherited DNA mismatch repair defects, causing variable number tandem repeat (VNTR) length alterations.

Page 18: OncoPrint Profiling of Colorectal Tumor Suppressors

• Visual mutation profiling in colorectal adenocarcinoma shows alteration frequencies: BMPR2 ($9\%$), BRCA2 ($7\%$), AKAP12 ($6\%$), AXIN1 ($4\%$), BRCA1 ($4\%$), BRMS1 ($1.3\%$), BMP2 ($1.1\%$), and BMPR1B ($1\%$).

Page 19: Patient Cohort Sorting by MSI Status

• Tracking individual patient identifiers (such as coadread_dfci_2016_407 and coadread_dfci_2016_3535) arranged by microsatellite instability status (MSI-high vs. MSS) to assess mutation patterns.

Page 20: Mapping Somatic Mutations Across MSI Cohorts

• Somatic variant calls (e.g., BRCA1 P145L, BMP2 A368V, BMPR2 Y247C, AXIN1 R395C) can be correlated with MSI classification to evaluate whether specific lesions are enriched in MSI-high tumors.

Page 21: Clinical Oncology Application: Melanoma Cohort Analysis

• Clinical investigation utilizing the Skin Cutaneous Melanoma (SKCM) TCGA provisional cohort to evaluate BRAF mutations and patient survival.

Page 22: BRAF Mutation Prevalence in Cutaneous Melanoma

• Somatic BRAF alterations occur in $52\%$ of evaluated SKCM cases ($366$ patients / $368$ samples), primarily consisting of putative driver missense mutations.

Page 23: Kaplan-Meier Overall Survival for BRAF Alterations

• Melanoma patients harboring BRAF alterations exhibit significantly altered overall survival curves compared to cases without alterations (Logrank test $p = 0.0315$).

Page 24: Comprehensive Oncoprotein Query Panel

• A panel of $16$ oncoproteins for mutational profiling in SKCM: ACVR1, ALK, ARAF, BRAF, CTNNB1, EGFR, FGFR2, FLT3, FRK, HRAS, JAK2, KRAS, MTOR, NRAS, PRKACA, and RAF1.

Page 25: Multi-Gene OncoPrint Profiling in Melanoma

• Mutation frequencies across the $16$-oncoprotein set in SKCM: BRAF ($52\%$), NRAS ($27\%$), ALK ($9\%$), FLT3 ($9\%$), FGFR2 ($8\%$), EGFR ($7\%$), MTOR ($7\%$), CTNNB1 ($5\%$), RAF1 ($3\%$), FRK ($2.2\%$), JAK2 ($2.2\%$), KRAS ($1.9\%$), ARAF ($1.6\%$), HRAS ($1.4\%$), ACVR1 ($1.1\%$), and PRKACA ($1.1\%$).

Page 26: Survival Dynamics Across the Oncoprotein Panel

• Patients with an alteration in any gene within the $16$-oncoprotein set exhibit significantly increased overall survival compared to patients lacking alterations (Logrank test $p = 0.0140$).

Page 27: Biological Mechanism: Transactivation of Apoptosis-Effector Genes

• Proliferation genes and apoptosis genes share common upstream transactivators.
• An increased burden of oncoprotein coding region mutations correlates with higher expression of apoptosis-effector genes, resulting in lower tumor lethality and improved clinical outcomes (Yavorski & Blanck, 2016).

Page 28: Characteristics of Unprocessed Sequencing Data

• Direct output reads from sequencing instruments representing tumors and matched normal tissues (e.g., blood).
• Whole genome sequences (WGS), which lack processed mutation datasets for either tumor or normal specimens.
• Patient-matched normal control sequences (exomes, WGS, RNASeq).

Page 29: Genomic Data Commons (GDC) Portal Navigation

• Primary web portal: gdc.cancer.gov / gdc-portal.nci.nih.gov.
• Provides tools to filter cohort files and generate manifest files for bulk authorized downloads.

Page 30: Raw Sequence BAM File Metadata

• Representative metadata for controlled raw aligned exome reads: UUID 50ffb480-b361-4231-8766-525b964e41ba, project TCGA-BRCA, format BAM, size $17.54\,\text{GB}$, experimental strategy WXS, platform Illumina.

Page 31: Research Applications of Unprocessed Sequencing Reads

• Detection of structural variants and non-standard copy number variations.
• Identification of somatic variants occurring in histologically normal host tissues.

Page 32: Direct Extraction of Raw Sequence Reads

• Raw sequencing files yield reads of approximately $100\,\text{nucleotides}$.
• Custom scripts (developed by Tommy Gill and Mohammad Samy) scan raw reads for specific target sequences and output matched reads to tabular files.

Page 33: Applied Research: Mining Controlled Genomic Repositories

• Practical project: Secondary mining of controlled-access raw exome datasets to identify immunological markers.

Page 34: Mechanism of Somatic V(D)J Recombination

• Somatic recombination rearranges discrete Variable (V), Diversity (D), and Joining (J) segments during lymphocyte development, creating diverse antigen receptors in B and T cells.

Page 35: Repertoire of Human Immune Receptor Loci

• Seven immune receptor gene loci exist in humans:
  • Three loci encoding antibodies (immunoglobulins).
  • Two loci encoding gamma-delta ($\gamma\delta$) T-cell receptors.
  • Two loci encoding alpha-beta ($\alpha\beta$) T-cell receptors (TRA and TRB).

Page 36: Functional Roles of Alpha-Beta T-Cell Receptors

• Represent the most abundant and clinically characterized T-cell subset.
• Recognize peptide antigens, including mutant tumor neoantigens.
• The TRB chain plays the primary role in antigen binding within the TRA/TRB heterodimer.

Page 37: Detection of Infiltrating Immune Receptors in Tumor Exomes

• Tumor biopsy tissue contains microenvironment stromal and tumor-infiltrating lymphocytes (TILs).
• Although the bulk of WXS reads derive from tumor cells, recombined V(D)J sequences from infiltrating immune cells are detectable within raw exome data.

Page 38: Bioinformatic Pipeline for V(D)J Identification

• Step 1: Low-stringency substring search across reads targeting known V- and J-region segments.
• Step 2: High-stringency alignment against the IMGT database to validate true V(D)J recombination usage.

Page 39: Target Nucleotide Substrings for V and J Matching

• Example $10$-nucleotide V-region strings: TGAATACTTC, ATACTTCCAG, CTTCCAGCAC, CCAGCACTGG, AGGTGCTGGG, TGCTGGGGGT, TGGGGGTCCC, GGGTCCCCTG, ACTGGTACTT, GGTACTTCGA.
• Example $10$-nucleotide J-region strings: TCTCGCACAG, CGCACAGTAA, ACAGTAATAC, GTAATACACG, GTAATACACA, TTTACAACAT, ACAACATTGA, ACATTGATTC, TTGATTCGGA, TAAAATCCTG.

Page 40: Read Length Constraints and V-J Junction Spanning

• Recombination places the selected V and J segments directly adjacent to one another.
• A single high-throughput sequencing read of approximately $75\,\text{nucleotides}$ is long enough to span the joined V-J junction.

Page 41: Addressing Genomic Non-Specific Matches

• Many non-recombinant genomic reads match short $10$-nucleotide V or J search strings by chance.
• Confirmation requires rigorous secondary alignment to verified IMGT germline reference sequences.

Page 42: Alignment Verification with IMGT Reference Standards

• Target reads (such as TRA rearrangements) are aligned against IMGT reference standards (www.imgt.org) to confirm segment identity and junction authenticity.

Page 43: Tabulated CDR3 Repertoires from Whole Exome Sequencing

• Custom processing of TCGA tumor exomes yields large tables ($> 7,500$ entries) cataloging V-segment ID, J-segment ID, alignment lengths, match percentages, and deduced CDR3 amino acid sequences.

Page 44: Clinical Prognostic Value of Immune Repertoire Mining

• Circulating blood-based antiviral T-cell receptor CDR3 sequences correlate with worse overall survival in neuroblastoma (Kacsoh et al., 2023).
• Chemical complementarity between tumor-resident TCR CDR3 sequences and MAGEA3/MAGEA6 tumor antigens correlates with increased survival in melanoma (Eakins et al.).

Page 45: Genomic Database Architecture Summary

• US disease-related genomic databases are centralized under the NIH and hosted through dbGaP.
• dbGaP provides a publicly visible study catalog, but access to raw sequence reads requires NIH Data Access Committee approval.
• Processed TCGA mutation, RNASeq, and clinical datasets are open to the public via cBioPortal.
• Complementary databases are maintained across Europe (COSMIC, ENA, EGA, UK Biobank) and China (CGGA).

Page 46: Analytical Value of Mining Raw Sequencing Data

• Direct access to raw sequence data permits exploratory mining for features outside the primary scope of the original study.
• Mining solid tumor exome files for infiltrating immune receptor recombinations provides clinically relevant prognostic indicators.