Error Rates for Nanopore Discrimination Among Cytosine Markers Along Individual DNA Strands
Engineering
BME 195
A recent discovery from the Human Genome Project is that humans have significantly fewer genes than expected, numbering approximately 20,000. To make up for the low number, these genes have the capacity to be regulated, allowing cells to specialize despite carrying the same genetic information by turning on and off genes which are needed for its function. This specialization is made possible by markers placed on DNA bases that regulate the expression of the nearby gene. Current methods of detecting these markers require extensive chemical preparation which is costly, error prone, and only resolve if the DNA has been marked, and not which marker is present, which can be misleading because different markers can have opposite effects on the expression of genes. I have shown that the nanopore system can distinguish between no marker, a marker which lowers gene expression, and a marker which occasionally raises gene expression, without the need for any chemical preparation at all. By combining the signals produced from the nanopore setup with various machine learning algorithms, my work has shown that, based on the surrounding DNA context, error rates ranging from 1.8% to 12.3% are achieved in discrimination between the three possibilities on a single molecule of DNA. The paper containing these results is in progress, and expected to be submitted to the Journal of the American Chemical Society (JACS) within the month.
