The profile Hidden Markov model (HMM) is a statistical
model
of a multiple sequence alignment (MSA).  Profile HMMs
are used in detecting sequence membership of protein families, protein
structure prediction, and sequence alignment.  
However, on evolutionarily divergent datasets, the accuracy
of alignments produced using profile HMMs degrade.  
We present a new statistical model for
representing  an MSA using multiple HMMs, which we call the Family of
HMMs (fHMM).  
We show that using many HMMs instead of a single HMM
to represent an MSA produces better alignment accuracy 
for sequence
datasets with high evolutionary diameter.  
We present TIPP, a new method for  
taxonomic identification and profiling, that uses
the fHMM model.
 We show that TIPP results in better
classification accuracy (especially of novel reads) and
more accurate abundance profiling.   Furthermore, TIPP is more robust
to sequencing errors, especially those resulting from indels, 
than  other profiling methods.