Hi Travis (and all),
Knome's Director of Research here. As such, I'm not a client, so will leave it to others to speak from that viewpoint. But to fill in a bit on what we do, running securely on the Cloud, kGAP aligns, assembles, and calls variants from the raw data generated for a work order; richly annotates each resulting genome/exome; and wraps them into a jointly queriable dataset, which we deliver with simple software tools for quickly and flexibly browsing the data. Our website videos ('2 Minute Genomes') show how to use those tools.
As part of the service, we deliver our own shortlists of suspect variants, genes, and gene sets, to help clients make sense of the data and follow up empirically. To make those lists, we filter variants by sensible criteria such as functional class (effects on protein sequence and, by predictive algorithms, function), distribution among cases/controls/reference populations (per expectations of rarity/novelty, inheritance mode, penetrance, etc.), and, as appropriate, gene function (including reported site- and gene-associated phenotypes). For the last, we leverage project-driven curation: while samples are being sequenced, we curate project-relevant publications, to ensure that our database of genotype-phenotype associations is up-to-date when we get the data back and bounce each genome off our reference data. Our curators work in Bangalore (as suggested in Arvktr's answer), and our Cambridge team builds and runs our pipeline and end-user tools, analyzes all project-generated data, and handles all other tasks.
Hope that helps a bit. Happy to answer any further questions: npearson [at] knome.com.
Director of Research