As a teaching assistant in Brown’s first year seminar “Phage Hunters” I lead several freshman biology and computer science students in an independent bioinformatics research project. We began the semester looking for evidence of CRISPR protospacers in mycobateriophage genomes. The idea was to use blast and other tools to get students introduced to the bioinformatics investigation process. We covered the basics of the CRISPR/Cas system, wrote a python script to download genome sequences from phagesdb.org, and made a local blast database on Brown’s computer cluster.
Things were going well with the project, but a few weeks in I was having doubts as to how statistically valid our protospacer predictions were. Then, I re-read a paper by one of the leaders in the field and discovered a) they had already looked for protospacers, and b) found no conclusive evidence in mycobacteriophages. The author of the paper was also going to be at the SEA-PHAGES symposium we were planning to present our class results at, so that really spelled the end of the CRISPR project. We needed a new idea though – the course instructors were counting on the bioinformatics team to generate some research we could bring to the symposium. My solution: frantic searching on Google Scholar for anything relevant to bioinformatics and bacteriophages.
Within a few minutes I came upon a paper (1) that looked at the the usage of tetranucleotides in viral and bacterial genomes. The idea is that closely related genomes have similar signals in terms of tetranucelotide usage, and this signal can be used to look at relationships independent of alignment-based techniques. I had found a new idea for the project! This kind of analysis was also perfect for teaching bioinformatics. It introduces a lot of the concepts and language used in the field, like kmer counting and normalization. It is fairly straightforward to program, easy to apply to bacteriophage genomes and doesn’t require complicated statistics in a first level investigation.
I ran with this idea for the bioinformatics project and the results were quite exciting. We found tetranucleotide usage was well conserved within mycobacteriophage cluster (a way to group phage based on pariwise nucleotide alignment and gene content comparisons) and divergent between clusters. We built phylogenetic trees that closely corresponded to published trees, looked for horizontal gene transfer and were able to accurately cluster unknown phage – all based on the usage of 4-letter words within the genomes. For a more detailed overview of the work, check out the abstract I submitted for the International Society for Computational Biology Student Council conference.
One of the first year students, Chen Ye, and I are also going to be presenting this research at the SEA-PHAGES symposium at HHMI’s Janelia Farm this weekend. Check back for an update with our poster and other thoughts from the conference!
1. Pride, D.T., Wassenaar, T.M., Ghose, C., and Blaser, M.J. (2006). Evidence of host-virus co-evolution in tetranucleotide usage patterns of bacteriophages and eukaryotic viruses. BMC Genomics 7, 8.