Several statistical models for calling the consensus of a slice are available within libSlice. By default, the consensus is called accoring to the Conic Ambiguity Model, which computes the consensus based on the cumulative quality value for each component in relation to the total quality value for the entire slice. Details on all of the model are available on the Understanding Ambiguity Codes page under "Additional Information".
The consensus quality value is calculated using the procedure described in Churchill, G.A. and Waterman, M.S. "The accuracy of DNA sequences: Estimating sequence quality." Genomics 14, pp. 89-98 (1992). The essential step in the calculation is to use Bayes rule to calculate for each of A,C,G,T and gap, five probabilities of error that the given base is the consensus, given the quality values in the slice. Each indicates the probability of error that the associated base is the consensus of the slice. The maximum value is called the consensus quality value and represents the probability that the unambiguous consensus is correct.
The attributes considered include if there are conflicting reads against the consensus, if there are reads in both the 3' and 5' directions, if the consensus is ambiguous, and the quality of the individual reads. The single number quality class designation combines all of these attributes to give an estimate as to the quality of the slice. This number can then be used to mark possible sites for editing or resequencing. More information is available on the Understanding Quality Classes page under "Additional Information".
$Date: 2005/07/29 20:40:45 $