Sharing:TocliniciansConsents in research & carePublication of research findingsAmong clinicians: means & formats of exchange, EHR integrationReturn of results
Reporting no-calls has low burden of interpretation: is site already annotated, or not?
Last, a warning: Let’s not dither, lest nationalized systems pass us.Biggest seq efforts already in such states, where individual and population-scale interests more directy align, especially where burdened with recessive disease &c.
• Conveying how bench and drylab methods work, fail, and change• Setting standards for pipeline validation, reporting, other sharing, and payment• Starting to teach med students• Convening stakeholders• Blunting hypeProprietary and ConfidentialWhat we, as a community, are doing right
• Plain English (see genomena.com/variants)• Treatment response findings• Reference genome(s)• Genotypes vs. variants• Sharing individuated genomesProprietary and ConfidentialWhat we, as a community, are neglecting
• Read: accurately sequence person’s chromosomesRecommend: Align & call against a genome most like this person’sWhere hard to guess (e.g., HLA), try many• Write: compress person’s genome; compare to others’ via common coordinatesRecommend: Report against human ancestral referenceReport no-calls• Interpret: understand person who carries this genomeRecommend: Focus on genotypes, not variantsFor QC, start with heterozygosityProprietary and ConfidentialReference genome: 3 distinct uses
• Arbitrary, unrealistic, and ethnocentric• Misaligns most real people’s genomes• Hides informative summary patterns (QC, functionally relevant evolution)• Includes rare and risky variants• Mismatches gene-specific references used by clinical geneticistsProprietary and ConfidentialReference genome: Why today’s failsSwitch to common-variant-only reference?Switch to healthy-variant-only reference?Instead…Switch to common-variant-only reference? What’s common varies.Switch to healthy-variant-only reference? What’s healthy depends.
• Read: accurately sequence person’s chromosomesRecommend: Align & call against a genome most like this person’s.Where hard to guess (e.g., HLA), try many.• Write: compress person’s genome; compare to others’ via common coordinatesRecommend: Report against human ancestral reference.Report no-calls.• Interpret: understand person who carries this genomeRecommend: Focus on genotypes, not variants.For QC, start with heterozygosity.Proprietary and ConfidentialReference genome: 3 distinct uses
• At each site, base carried by last forebear of all people• Like current reference, comprises mostly common and healthy variants• Includes source DNA for nearly all alignment-relevant chunks of real genomes• Roughly equidistant from everyone• Clearly reveals summary patterns of variation and evolution• Lesson long learned for mtDNA1But gene-specific reference discrepancies will remain, so…1See Behar et al. 2012 (PMID 22482806), as well as Balasubramanian et al. 2011 (PMID 21205862)Proprietary and ConfidentialReport against human ancestral reference
• Reference-independent• More clearly convey what’s risky & what’s not, in given person• Slightly bigger data• Needed, to capture complexity that big studies are already revealingNeatly convey: dominant/recessive/complex site-specific effectssex-specific risk (sex chromosome epistasis)Readily extrapolate to: classic compound het etiologyhaplotypes (and diplotypes)other compound etiologyProprietary and ConfidentialClassify genotypes, not variants
• What variants appear together, at what zygosities, in sick vs. healthy genomes?• Needed, to capture complex etiologyClassic compound heterozygous etiologySex-specific risk (sex chromosome epistasis)Other compound association (e.g., classic burden)• Lets us refine sequence (phase, impute)Proprietary and ConfidentialQuery individuated genomes, not allele frequencies
• Defined by multiple variants (rs429358C, rs7412C)• One variant rare, one common (neither is a mutation)• Harmful for Alzheimer disease & longevity…but helpful for cancer?• Genotype matters…as does interaction (e.g., intronic BACE1 variant)• Chronically hard to call…highlights need to report no-calls• Other familiar examples: globinopathies, BRCA1 modifiers, &c.Proprietary and ConfidentialWhy this stuff is tricky: APOE4 example
• Secure (HIPAA-, Safe Harbor-compliant) web platform for interpreting called humangenomes• Smart interface to flexibly annotate & compare genomes, to shortlist candidatevariants, genes, & gene sets• Leverages field’s deepest functional knowledge base, with rigorous clinical-depthcuration of published findings, well structured ontology, and smart interaction modeling• Statistically robust methods for interpreting single/multi-proband, matched tumor, (multi-)kindred, and big case/control cohort studies• Suits central labs’ needs to manage clients’ human genome data, help interpret it, andbroker sensible sharing• Clinicians, stay tuned.Proprietary and ConfidentialIngenuity® Variant Analysis™ in a nutshell