This method can be used when simple, unquantified, base call quality values are available. Instead of simply counting base type frequencies it sums the quality values. Hence a column of 4 bases A, A, A and T with confidence values 10, 10, 10 and 50 would give combined totals of 30/80 for A and 50/80 for T (compared to 3/4 for A and 1/4 for T when using frequencies). As with the unweighted frequency method this sets the confidence value of the consensus base to be the the fraction of the chosen base type weights over the total weights (62.5 in the above example).

The quality cutoff parameter controls which bases are used in the calculation. Only bases with quality values greater than or equal to the quality cutoff are used, otherwise they are completely ignored and have no effect on either the base type chosen for the consensus or the consensus confidence value. In the above example setting the quality cutoff to 20 would give a T with confidence 100 (100 * 50/50).

In the event that more than one base type is calculated to have the same weight, and this exceeds the consensus cutoff, the bases are assigned in descending order of precedence: A, C, G and T.

This is Rule IV of Bonfield,J.K. and Staden,R. The application of numerical estimates of base calling accuracy to DNA sequencing projects. Nucleic Acids Research 23, 1406-1410 (1995).

URL: http://www.mrc-lmb.cam.ac.uk/pubseq/manual/gap4_unix_119.html