|
Clustering is a popular method utilized in analyzing T-RFLP data that is used in studying some
human microbial ecosystems (Zhou et al. 2007 for example). Clustering is used to identify microbial community kinds that share similar structure.
Clustering T-RFLP data takes place in two steps: 1) similar operational taxonomic units (OTUs) between different samples are identified based on
the similarity of associated fragment lengths (a step referred to as alignment or binning) and 2) different sample profiles are clustered based
on the abundance of the identified OTU's in each sample. Both steps jointly affect the accuracy of the final clustering. In this talk I present
a new, model-based-Bayesian-simultaneous alignment approach to inferring the relationships between sampled T-RFLP profiles. I compare this method to a distance based alignment approach that we developed previously. I utilize simulated data in these comparisons. Preliminary results indicate that the new model-based approach is superior to distance-based one in improving the recovery of the true clustering structure of the data
All interested faculty, staff, and graduate students are invited to attend.
|
|