5 ms·
Interesting to think about how you can ensure your anonymity and still take advantage of a cheap genome!! Some preliminary ideas: - misidentify yourself (eg g
by Gatsky 8y ago
Interesting to think about how you can ensure your anonymity and still take advantage of a cheap genome!!
Some preliminary ideas:
- misidentify yourself (eg get a few friends and submit each other's samples)
- mix your DNA with a mammalian organism (something like a mouse). It will get flagged as a bad sample by them probably, due to poor alignment metrics and too many variants, but with the knowledge of which DNA you put in you can do a dual alignment to eg. the mouse and human genome and purify the reads that came from you (there are published methods to do this kind of thing). Not sure how you would get an equimolar mix of mouse and your own DNA, but for $200 you could submit multiple ratios I guess.
- enzymatically modify your DNA. There are some enzymes which cause characteristic mutations. They will turn up as real variants in their pipeline, but you can use the sequence context to filter them out. Kind of like chemical code obfuscation. This would mean extracting your own DNA and treating it, probably not that feasible.
- nosuchthing 8y agoThis project was made exactly for what you’re describing: http://biononymous.me/ http://biononymous.me/
- Gatsky 8y agoInteresting. To be honest, if someone is able to collect intact tissue from you without your knowledge, there is no way to hide. If you can control the donation then it is different.
- deytempo 8y ago1) yes 2) no 3) no The first one is the best because if enough people did it, it would create enough doubt in the accuracy of the stored datasets that they wouldn’t be worth anything to anyone except the person who owns the DNA. The problem is, as this becomes more popular more people related to you are going to get it done, meaning they will be able to automatically fit you into an identity based on who else had it done, their known identities and just the similarities between their DNA and yours. Trying to maintain anonymity only works if the data set is small. It looks like there is a market for a completely discrete secure dna sequencing company that doesn’t keep a copy of your genome after the fact and sends you a copy that is encrypted. Also it might be feasible to have some kind of kit that divides up the chromosomes into groups to be sent to different labs. It can’t be that difficult and expensive to do this. Why not just wait until there is a kit you can buy to do it directly from your PC if you are concerned with privacy?
- Gatsky 8y agoI wouldn’t rule out 2. It works quite well, the dual alignment method is used in the research setting for human tumours grown in mice for example. Of course, a very motivated adversary would work this out, but you could use more unusual organisms if you have a reference genome for them. The problem with 1 is that there may be significant downsides to misidentifying your genome eg you swapped with someone who commits a crime.
- klmr 8y agoIdentifying and removing contaminants is a routine step in DNA-seq analysis. They might still flag and discard the sample internally but treating it would be trivial.
- Real_S 8y ago[2] wont work on a whole genome because of the many differences present between mammals after millions of years of evolution. The "alignment" of a small sequence may be confused between human and mouse, but not longer sequences and definitely not whole genomes (alignment using a tool like BLAST or BLAT). I wrote a summary of genetic information discussing this very idea. See the PDF under "What is Genetic Info?", read the Phylogenetics section: https://www.geneinfosec.com/more-info/ https://www.geneinfosec.com/more-info/
- Gatsky 8y agoNo this is not correct, look up modern whole genome pipelines. Standard read lengths are 75-150 base pairs. Whole genome alignment has never been done with BLAST or BLAT (this would take an extremely long time). BWA is most commonly used. Running the standard data pipeline on a mouse human mixture will cause many mouse sequences to be aligned to the human genome.
- Real_S 8y agoMaybe not exactly something you can do right now, but my invention of molecular cryptography will allow genomes to be sequenced without revealing identity or health information. You are somewhat close with [3] geneinfosec.com Also note that DNA is readily identifiable, so misidentifying yourself does not work.