A consensus based on mapped reads cannot be generated in regions that meet or are below the value set for the low coverage threshold, there are several options for handling these low coverage regions:
- Remove regions with low coverage. When using this option, no consensus sequence is created for the low coverage regions. There are two ways of creating the consensus sequence from the remaining contiguous stretches of high coverage: either the consensus sequence is split into separate sequence when there is a low coverage region, or the low coverage region is simply ignored, and the high-coverage regions are directly joined (in this case, an annotation is added at the position where a low coverage region is removed in the consensus sequence produced, see below).
- Insert 'N' ambiguity symbols. This will simply add Ns for each base in the low coverage region. An annotation is added for the low coverage region in the consensus sequence produced (see below).
- Fill from reference sequence. This option will use the sequence from the reference to construct the consensus sequence for low coverage regions. An annotation is added for the low coverage region in the consensus sequence produced (see below).
In addition to deciding how to handle low coverage regions, you can also decide how to handle conflicts or disagreement between the reads when building a consensus sequence in regions above the low coverage threshold:
- Vote. Whenever the reads disagree on the base at a given position, the vote resolution will let the majority of the reads decide which base is correct. In addition, you can specify to let the voting use the base calling quality scores from the reads. This is done be simply adding all quality scores for each base and let the sum determine which one is correct. The base with the highest total quality scores will be chosen. If there are two bases that end up summing to the same total quality score for all reads at that location, A is preferred before C, C before G, and G before T. An annotation with the complete information that was used to resolve the conflict will be added.
- Insert ambiguity codes. When this options is selected, read conflicts are addressed by using an ambiguity code representing all read bases represented at the reference location. The problem with the voting option is that it will not be able to represent true biological heterozygous variation in the data. For a diploid genome, if two different alleles are present in an almost even number of reads, only one will be represented in the consensus sequence. With the option to insert ambiguity codes, this can be solved. However, if an ambiguity code would always be inserted if just one read had a different base, there would be an ambiguity code whenever there was a sequencing error. In high-coverage NGS data that would be a big problem, because sequencing errors would be abundant. To solve this problem, you can specify a Noise threshold. The default value for this is 0.1 which means that for a base to contribute to the ambiguity code, it must be in at least 10 % of the reads at a given position. The Minimum nucleotide count specifies the minimum number of reads that are required before a nucleotide is included. Nucleotides below this limit are considered noise. If no nucleotide passes the noise filters, the position is omitted from the consensus sequence.
- Use quality score. The "Use quality score" checkbox option is available for conflicts regardless of whether "Vote" or "Insert ambiguity codes" has been selected. The "Use quality score" checkbox option allows you to use the base calling quality scores from the reads. This is done by simply adding all the quality scores for each base and let the sum determine which bases to consider. In other words, if quality scores are used, we will sum the quality score (instead of amount of reads) for each base on each position before applying the noise filters and finally call the consensus symbol.
Click Next to set the output option as shown in figure 23.13).
The annotations that can be added to the consensus sequence produced by this tool show both conflicts that have been resolved and low coverage regions (unless you have chosen to split the consensus sequence). Please note that for large data sets, this can amount to a very high number of annotations, which will cause the tool to take longer to complete, and the result will take up much more disk space.
It is also possible to transfer existing annotations to the consensus sequence produced. Please note that since the consensus sequence produced may be broken up, the annotations will also be broken up, and you cannot expect them to have the same length as before. In some cases, gaps and low-coverage regions will lead to differences in the sequence coordinates between the input data and the new consensus sequence. The annotations copied will be placed in the region on the consensus that corresponds to the region on the input data, but the actual coordinates might have changed.
Copied/transferred annotations will contain the same qualifier text as the original. That is, the text is not updated. As an example, if the annotation contains 'translation' as qualifier text this translation will be copied to the new sequence and will thus reflect the translation of the original sequence, not the new sequence, which may differ.
The resulting consensus sequence (or sequences) will have quality scores assigned if quality scores were found in the reads used to call the consensus. For a given consensus symbol we compute its quality score from the "column" in the read mapping. Let be the sum of all quality scores corresponding to the "column" below , and let be the sum of all quality scores from that column that supported 23.1. Let , then we will assign the quality score of where
- By supporting a consensus symbol, we understand the following: when conflicts are resolved using voting, then only the reads having the symbol that is eventually called are said to support the consensus. When ambiguity codes are used instead, all reads contribute to the called consensus and thus .