Thread by @AlbertVilella

Is @NebulaGenomics using @MGI_BGI for their 30x 2x150bp #WGS genome sequencing?

I received an email from @NebulaGenomics today, with a discount code for a $299 genome sequencing offer. Which made me wonder: are they using the already cheaper @MGI_BGI #DNBSEQTx sequencing for this? It would mean they can run it at cost or make a small profit from it.

Then you realise there is "analysis" charge on top of the $299 which you can't untick, so the $299 #WGS 30x genome goes up to $499. Still, at $499, you would find it difficult to make money if you were @NebulaGenomics and you were using something like an @illumina #NovaSeq

as the list price for a genome on each #NovaSeq #S4 kit is around $600. So the only option to be able to run this and not make a loss is to go for @MGI_BGI #DNBSEQTx: send 10E5-10E6 samples to MGI Shenzhen, and have them sequenced on their factory-like Tx system:

Would @NebulaGenomics have to disclose that their samples are being sent to MGI Shenzhen for sequencing? They are the company that pioneered Homomorphic Encryption for genomic data, so in terms of data privacy, there shouldn't be a *technical* reason to complain,

but maybe there is a compliance requirement for biological/DNA samples being sent, say, from the US to China regardless of how they are used or profiled. Happy to learn more on this, if someone knows the details.

The other option, and this wouldn't be a surprise having heard the way George Church describes the @NebulaGenomics model in interviews, is that they are making a small loss for every genome, but the homomorphic encryption method (#crypto #genomics) that they are using is building

a large #WGS database that can be mined for new drug targets, low-frequency variant identification or similar purposes. This large database, which is inherently encrypted at source, would allow @NebulaGenomics to follow the #23andMe model of reselling the data to large parties.

The difference here is that @23andMe only (or mostly) has profiles for around 1M common SNPs, and can't give any information on low-frequency variants or private SNPs, as one can do with 30x #WGS approaches.