After Deduplication, a consensus sequence is derived to determine the most probable CDR3 representation from each read cluster. To derive consensus, quality values are used to determine the most probable sequence at each position of the CDR3. Since quality values are simply a transformation of the p-values, all matching base calls at a given position can be combined together using Fisher's method. The combined p-values for each base call are compared, the base with the lowest combined p-value (highest combined quality) is reported as the base call for each postion in the read. A consensus quality is also defined by transforming the combined p value to a quality score for each position. Any consensus quality scores that exceed 60 are capped at 60, since we are limited in the number of ASCII characters for reported.
Using this method, a consensus sequence and consensus quality string are derived for both the UMI and CDR3 sequences for each UMI family.