| | This method was based on comparison of ESTs from the same gene. Clustered ESTs |
| | were directly downloaded from UniGene. To correct possible misclustering, UniGe |
| | ne clusters were mapped onto genomic sequence by homology search followed by dy |
| | namic programming. Many clustering errors (e.g. mixing up similar ESTs from par |
| | alogous genes) were cleaned by screening of the ESTs within each cluster agains |
| | t the genomic sequence. SNPs were detected by applying Bayesian statistical met |
| | hods to these corrected EST clusters. Briefly, Trace chromatogram data of EST s |
| | equences in Unigene were processed with PHRED. To identify likely SNPs, single |
| | base mismatches were reported from multiple sequence alignments produced by the |
| | programs BRO and POA for each Unigene cluster. BRO corrected possible misrepor |
| | ted EST orientations, while POA identified and analyzed non-linear alignment st |
| | ructures indicative of gene mixing/chimeras that might produce spurious SNPs. B |
| | ayesian inference was used to weigh evidence for true polymorphism versus seque |
| | ncing error, misalignment or ambiguity, misclustering or chimeric EST sequences |
| | , assessing data such as raw chromatogram height, sharpness, overlap and spacin |
| | g; sequencing error rates; context-sensitivity; cDNA library origin etc. |
