Protein Family Classi cation using Sparse Markov Transducers

In this paper we present a method for classifying proteins into families using sparse Markov transducers (SMTs). Sparse Markov transducers, similar to probabilistic suÆx trees, estimate a probability distribution conditioned on an input sequence. SMTs generalize probabilistic suÆx trees by allowing for wild-cards in the conditioning sequences. Because substitutions of amino acids are common in protein families, incorporating wildcards into the model signi cantly improves classi cation performance. We present two models for building protein family classi ers using SMTs. We also present eÆcient data structures to improve the memory usage of the models. We evaluate SMTs by building protein family classi ers using the Pfam database and compare our results to previously published results.