Other apparently identical isolates likely represented chance matches rather than clonal genotypes, as they represented pairs of isolates from distant areas and all of these matches were samples that had one or more non-amplifying alleles.
Because this is a much greater concentration of birth dates than actually happens for those aged over 65, this assumption leads to over-estimation of the number of chance matches and therefore the corresponding false match rate.
In total, 95.5% of these 149 peptides align to the genome only once (in the original predicted locus), indicating that the majority are indeed unique and not chance matches to another mis-predicted sequence.
Although localized sequence similarity can indicate common function, care must be taken to distinguish between sequence similarity arising from functional constraints within folds and similarity arising from chance matches in short sequence stretches.
Additionally, substantially more statistical testing would be needed in order to rule out the possibility of chance matches in the binding site, and to show a specific association between transcripts containing the binding site and transcripts that are upregulated.
Whether or not such chance matches make up a sizeable proportion of all matches, and therefore affect the utility of the linked data, can be seen by comparing the number of chance (and therefore false) matches with the total number of achieved matches; that is, by estimating a false match rate per hospital separation date (termed the 'false match rate').
The input data for a Blastn run are: query sequence, sequence set to be searched, Expected number of chance matches in a random model (E-value), maximum number of hits to be displayed, maximum number of aligned sequences to be displayed, size of the string to be searched in pairwise comparison (word size), scoring parameters, and filtering and masking options.
Peptides shared between proteins arise from gene families with closely related sequences, the database storing different splice products of the same gene, database errors (e.g. multiple records for the same gene products) and chance matches of typically short peptides in otherwise unrelated proteins.
Such p-values are calculated as the probability that the descriptor combination in the left hand side of the rule by chance matches at least the observed number of genes in the class found on the right-hand side (see Additional file 2 for all rules and p-values).
In the absence of such information, the expected numbers of chance matches can be estimated by assuming that hospital separations and entries into residential aged care are independent Poisson processes and then substituting in the relevant expected daily numbers of hospital separations and residential aged care entry events.
A key factor determining whether or not the strategy could be used is the number of candidate matches being considered for an exit event: if this is too high the probability of chance matches becomes too great and the false match rate increases accordingly.
