Automated genotyping: combining neural networks and decision trees to perform robust allele calling

The Human Genome Project has revolutionized the field of modem genetics by pmwiding increasingly dense high resolution genetic mops of the human genome. In order to investigate inheritance patterns of genetic disorder, selected areas of the genome are genotyped using densely distributed genetic markers. Due to the complezity associated with most of these inheritance pattems hundreds of thousands of ezperiments are required to localize areas small enough to sequence in its entirety. Currently, the main bottleneck in high throughput genotyping is the amount of manual editing of allele calls required. Earlier attempts to automate allele calling have locked accuracy, consequently colling for manual inspection of up to 40% of the date. In addition, they tend to be computationally demonding and slow in execution. We present a solution which outperform a state of the art allele calling algorithms both in terms of speed and quality.