Blog @ Illumina
Real scientists. Real commentary.

Illumina Discovery Symposium at ASHG

Amy Cullinan, Ph.D.
| Nov 04, 2013

DNA strandOur third discovery symposium was held last month ahead of the American Society for Human Genetics Meeting in Boston. The day’s excellent series of speakers covered a range of genomic research topics from Drosophila transcriptomics, to placental metagenomics, to building a better reference genome.

Jon Sebat from the Departments of Psychiatry, and Cellular & Molecular Medicine at University of California at San Diego presented on the role of copy number variants (CNVs) in neuropsychiatric disorders. As the evidence for enriched mutational burden of CNVs in schizophrenia is well-established; Dr. Sebat presented an update from a mega-analysis in a very large schizophrenia cohort. Together with the psychiatric genomics consortium, comprising a dozen research universities and institutes, this comprehensive study has identified >100 variants associated with risk of schizophrenia risk, including an intronic variant in an interesting microRNA (MIR137). The next step in this research is to put all these and other variants onto an Illumina array, and test an even larger sample to replicate the findings.  

Manolis Kellis from the Broad Institute of MIT and Harvard and the MIT Computer Science & Artificial Intelligence Lab spoke on extensive systematic efforts aimed at understanding human variation and its role in disease. Dr. Kellis detailed several approaches toward relating genotype-tissue expression to disease, the most interesting of which was an effort to map the genotype and methylome of each of more than 700 Alzheimer’s patients in an EWAS, or epigenome-wide association study. The study found that most epigenomic variability is genotype-driven, with single nucleotide polymorphism (SNP) -associated CpGs depleted in promoters and significantly enriched in distal enhancers. Although massively integrated, systems-level studies of genomes and gene regulation have made great inroads in our understanding, Kellis emphasized that challenges remain in validating loci and unraveling mechanisms, pathways, and potential drug targets.

Nazneen Rahman, head of Cancer Genetics for London’s Institute of Cancer Research discussed many facets of clinical testing for cancer predisposition genes, and how the current model is not serving patients very well. Dr. Rahman talked about the goals of Mainstreaming Cancer Genetics (MCG), a Wellcome Trust-funded project aimed at developing the assays, informatics, infrastructure, education, ethics, and evaluation criteria to foster the adoption of germline cancer testing into routine clinical care. Describing the development of technical, analytical, and interpretation aspects of the MCG program, Rahman provided a compelling rationale for using the TruSight Cancer panel. Using a targeted approach on 575 samples, 99% of the probes showed good performance, with a median coverage of 472.5x and 100% coverage of BRCA1/2 variants, with 100% specificity and sensitivity. Rahman also highlighted the advantages of high-throughput capability: 192 samples per run can be performed on HiSeq 2500, translating to 576 samples per week.

Bob Daber, technical director of Clinical Genomics at the University of Pennsylvania’s Center for Personalized Diagnostics discussed the myriad operational variables that were considered when they built their clinical NGS assays. Beginning with specific challenges specific to using NGS in an oncology setting, including confidently detecting complex genomic profiles in heterogenous samples that are often damaged by fixatives, Dr. Daber walked the audience through clinical development and validation processes from sample to report. Predictably, challenges were apparent in sample acquisition— getting enough “amplifiable” material— and in tweaking the bioinformatics analysis to flag biological events that may have been missed. Looking back on eight months of experience since clinical launch, they’ve catalogued 879 unique variants in their knowledgebase, over half of which are associated with disease. The vast majority of these are SNVs, although Dr. Daber did finish his talk with a case report of detecting a complex BRAF cis mutation in a melanoma patient that was missed using traditional single-gene tests, and in which multidrug treatment was beginning to show promising results.

Mike Eberle from Illumina closed out the day with an update on Platinum Genomes to assess and improve variant calls by creating a training and assessment catalogue of very high confidence whole-genome calls. Using a large pedigree of 17 individuals, including 11 siblings, error detection is dramatically improved and single errors can be identified in >99.7% of the variant positions. Defining parental inheritance in each of the 11 children and calling variants using a number of approaches helps define truth variants where the genotypes are 100% consistent with transmission of the parental haplotypes. The whole-genome catalogue of “truth” SNPs enables better training of filters to significantly improve sensitivity with little loss in precision. Eberle discussed several current projects, including using long-read information for calling insertions and deletions to help improve sensitivity of indel callers, and using kmer genotyping and targeted read counts for clarifying difficult-to-call variants.

Thanks to all of our amazing speakers and participants who came to Boston for this successful Discovery Symposium!