Reconstruction of private genomes through reference-based genotype imputation.

Published in: Genome biology (2023)

Our results show that genomes in imputation server reference panels can be vulnerable to reconstruction, implying that additional safeguards may need to be considered. We suggest possible mitigation measures based on our findings. Our work illustrates the value of adversarial algorithms in uncovering new privacy risks to help inform the genomics community towards secure data sharing practices.

Keyphrases

healthcare
health information
big data
machine learning
primary care
climate change
mental health
social media
electronic health record
deep learning
single cell
human health
artificial intelligence
data analysis