Frameshift Correction in De Novo Assembled Transcriptome Data Using Peptide Data, Blast Sequence Alignments and Hidden Markov Models

摘要

Frameshift errors in de novo sequenced transcriptome data are difficult to find since not all transcripts are covered with publicly available and cu-rated reference data. One approach for finding frameshift errors are hidden markov models. HMMs are a widely used approach in bioinformatics to identify patterns in sequences such as coding or conserved regions [1]. However there haven't been made many efforts using HMMs for detecting frameshift errors in nucleotide sequences. Here we introduce an approach for frameshift correction using hidden markov models, blast alignment data and peptide data derived from mass spectro-metry [2]. This algorithm is implemented as pipeline with three correction steps for transcriptomic sequences. First, it employs peptide data for a preliminary correction, followed by aligning the investigated sequences against publicly available protein databases such as the SwissProt database utilizing the BlastX alignment tool. Finally, the resulting alignment files for correction are used for creating training data sets for the HMM using known coding and shifted areas on the sequence. The trained HMM is then used to perform the final identification.