4. Discussion
4.1 Comparing BWA aln and BWA mem to improve ancient DNA
mapping
BWA aln with the MS parameters has proved to be effective in
ancient DNA mapping by disabling the seed function and decreasing the
difference tolerance in mapping (18). However, we did not find
significant differences in CRT (F = 1.42, P = 0.2870)
between the Schubert’s method and the BWA mem algorithm. It
indicated that BWA mem with default parameters (BWA mem -k
19 -r 1.5) was able to perform ancient genome mapping as well as BWAaln with the MS parameter. Additionally, the seed-reseed-extend
strategy in BWA mem can help to accelerate the mapping process,
and it resulted in an 8.13-fold decrease of MT than the BWA alnalgorithm. Therefore, BWA mem can reduce the contamination rate
and improve the accuracy of ancient genome analysis while consuming a
short time to run the mapping process. Additionally, the CRT was
maintained at the same level among different samples, demonstrating the
universal property on ancient genome mapping.
Soft clipping is one of most important issue to consider when using the
BWA mem . To determine whether soft clipping could help decrease
the contamination rate, we counted all soft clipped reads. The results
showed that 7.9% mapped reads were soft clipped during the mapping and
6% soft-clipped reads contained C-to-T and/or G-to-A changes within
soft-clipped regions. In other words, only ~0.47%
(7.9%*0.6%) mapped reads with damaged patterns were soft clipped,
which was a small proportion when considering the large number of
damaged endogenous DNA. As to hard clipping, it’s a special kind of soft
clipping to mark the multiple mapping of a read. But only 0.0036%
mapped reads showed damaged pattern. Therefore, the soft clipping can
hardly make a big impact on the further filtering of endogenous DNA by
using the deamination characteristic. In summary, the BWA memperformed equally well as the BWA aln with MS parameters when
considering the contamination rate used in this study, but BWAmem costs less running time than that of the BWA alnmethod. If we took all the conditions into consideration, BWA memperformed better.
4.2 Exploring more accurate and effective mapping parameters of BWA