The probability p i that a single read in pool i is a copy of a mutant allele is given by: This probability can be interpreted as the probability that a mutant read is read correctly by the sequencer plus the probability that a wild-type read is read with an error, transforming it to a mutant read.
Therefore, for a simulation, the error pattern is read with an offset varying from at the first run and incremented by for each run and finally the results are evaluated over a set of runs, as recommended in [13].
Thus, even if a particular transcript is read out with a single error, the software can determine its correct identity, although RNAs containing two or more errors can be detected but no longer corrected, as it is no longer possible to unambiguously identify them.
If the loss of sequence data can be afforded, reads with high error rates (for example containing low base quality scores or ambiguous bases) and short reads should be filtered prior to using DeconSeq to ensure high accuracy of the contaminant classifications.
Raw reads were preprocessed to remove reads with errors (see Methods) and the remaining 200 million reads were aligned with the pig genome reference assembly (Build 8).
First, reads with errors were expected to appear at lower frequencies than reads without errors in the whole dataset and for each individual sample.
Thus, BFAST permits a larger fraction of reads with errors to be correctly aligned.
In contrast, because sequencing errors are distributed throughout the read with three possible wrong base substitutions, reads with errors that create false positive junctions tend to be scattered as single, incorrect alignments.
Since the same sequencing error is not expected to occur in both paired-end reads, this PELE filter eliminates reads with errors that occurred during sequencing.
When using longer reads with high error rates (5 20 kb read with 4%15%% error per base), phasing performance was substantially improved.
