Candidate k-mers are stored in a data structure such as a Hamming graph, which connects k-mers within a fixed distance, or a Bloom filter.
Recent approaches have taken steps in the right direction by looking together at all k-mers within a small Hamming neighborhood.
Any repetitive k-mers within a string are counted only once since only the unique counts are used to create the quotient.
In Hammer, we proposed the simple choice of the consensus string as a way to correct the k-mers within a cluster.
It is possible, however, to develop a more sophisticated approach that allows the identification of multiple correct k-mers within a cluster (Wijaya et al., 2009).
Stanley et al. analyzed the distribution of 1- to 4-mers within a wide variety of organisms and found that some tend to cluster within genomes (usually in non-coding regions) and others tend to "repel" each other [ 12, 13].
There is no dedicated solution in T-IDBA that solves the issue of erroneous k-mers within a component and methods for isolating components are not sensitive to low-expressed isoforms.
Newer platforms, such as the PacBio sequencer that has a large amount of indel errors, will also undoubtedly challenge Hammer's current approach; however, we believe that the idea of looking at all the k-mers within a single cluster jointly is a powerful one and will continue to be useful.
$$\end{document} p e f = 1 − L 1 − E k − 1 + L exp − L 1 − 1 + L exp − L − 1 − E k L exp − L 1 − E k 1 − 1 + L exp − L. Sequencing errors will generate false k-mers that will increase the apparent number of k-mers within a given species, n t *.
Also, each node in a splicing graph is evaluated instead of each exon, with each read that contains a k-mer within a node contributing to that node.
Therefore in the de Bruijn graph, sp-branches started from a common k-mer in similar subspecies usually have (converge to) another common k-mer within a short distance, while cr-branches started from a common k-mer in multiple species seldom have another common k-mer within a short distance.
