Combining Bacterial Fingerprints

<p>Since there are ten times more bacterial than human cells in your body, it is at least interesting to know which bacteria call you home, how they got there, and what they are doing. Unfortunately, less than two percent of all bacteria are known, and only about five percent can be grown in the lab. Only in the last few years has it been possible to perform a complete survey of the "human microbiome", as the microbial ecosystem that is you is called. One technique is to collect all the bacterial DNA from a site on or in the human body, then to sequence a “fingerprint” region which can distinguish one species from another, even when the species are unknown. We then cluster these fingerprint sequences in order to determine the number of bacterial species present and to infer their likely function from known species to which they are closely related. Current techniques use only a single fingerprint region, because different fingerprints have different, usually unknown, biases. We have developed a technique for combining DNA sequence data from multiple fingerprints. The technique borrows algorithms from graphics, statistical analysis, and phylogenetic estimation. The main problem we currently face is scaling the algorithm up to tens of millions of DNA sequences. Perhaps you will have some ideas on how to do this!