Hi, thanks for your interesting work. Your conclusion is really inspiring and it seems that privacy analysis of single-cell research should raise attention. However, I have a quick question about the meaning of phenotype here. Could we directly use a known phenotype-scRNA-seq datasets to train a model to predict phenotype? What is the difference between this classifier and the attacks you discussed? Should we treat phenotype as information we need to protect? I am a bit confused here because it seems that HIV status is something like disease state, which should be disclosed in most of the research.
Thanks a lot!
Hi, thanks for your interesting work. Your conclusion is really inspiring and it seems that privacy analysis of single-cell research should raise attention. However, I have a quick question about the meaning of phenotype here. Could we directly use a known phenotype-scRNA-seq datasets to train a model to predict phenotype? What is the difference between this classifier and the attacks you discussed? Should we treat phenotype as information we need to protect? I am a bit confused here because it seems that HIV status is something like disease state, which should be disclosed in most of the research.
Thanks a lot!