Identified SNPs can be analysed concurrently in a flexible and interactive manner with PanGEA. Two different SNP analysis options are available, SNP statistics may either be displayed for each individual SNP-site or for each individual database sequence. For example, PanGEA displays for each SNP-site, a number of informative characteristics such as the total coverage, the percentage of sense ESTs mapping to the SNP-site and the allele frequencies.

A subset of the SNPs can be delimited according to quality, transcription direction and frequency. The quality of a SNP is assessed by several criteria such as the minimum sequence quality at the SNP, the minimum sequence quality in the neighborhood of the SNP, the maximum number of low alignment quality token in the neighborhood of the SNP and the minimum distance from the alignment ends. PanGEA allows to assess the effect of altered parameters by displaying concurrently the initial data and the data generated with the user-specified settings.

Moreover, PanGEA estimates basic benchmarks, such as the sensitivity and specificity achieved with the user-specified settings. These benchmarks are assessed from the bias observed in the SNP analysis. As there is no biological reason why alleles of the reference sequences should be more frequent than any other alleles, it can be expected that the number of SNP-sites ('n_ref') in which the most frequent allele is the reference-sequence-allele should be roughly equal to the number of SNP-sites ('n_not-ref') in which the most frequent allele is any other allele. If however 'n_ref' dramatically deviates from 'n_not-ref' the SNP analysis is biased (for details see here). This bias may be especially apparent when sequencing errors preferentially occur in certain regions of a sequence, like it has been demonstrated for the 454-technology where sequencing errors frequently occur close to homopolymers. However, using these benchmarks the effect of the user-specified parameters can be interactively estimated. It may, for example, be assessed whether increasing demands on sequence quality are effective to reduce the SNP-bias and thus exclude putative sequencing errors from the SNP analysis. We however caution the users, as these benchmarks are solely based on statistical considerations the number of true and false SNP-sites may significantly deviate from this estimates, they should merely be used as a basic guideline to assess the effect of the active settings.