Schematic overview of the Tracembler algorithm. Tracembler accepts as input one or more user-supplied query sequences and parameter specifications. The query sequence(s) and associated parameters are submitted using the QBLAST URL API to NCBI  (1). Tracembler analyzes these results, and if there are new sequences matching at a significance level below the user supplied E-value parameter, these sequences are used as queries in a new BLAST search (2). One round consists of BLAST searches of all acceptable matching sequences from the previous round. This process is repeated in a recursive manner until either all matching sequences are exhausted, a user-defined maximum round of recursion is reached, or a user defined maximum total number of BLAST queries is reached. For the final set of sequences, quality score and mate-pair distance constraint information is retrieved from NCBI. These sequences are assembled using CAP3 (3). Finally, novel contigs are compared to the query sequences using BLAST for local alignment and GenomeThreader for spliced-alignment (4).
Dong et al. BMC Bioinformatics 2007 8:151 doi:10.1186/1471-2105-8-151