paperKB
coga / coga-kb
Help
Sign in

Chunk #56 — Findings — Performance comparisons — Identity-by-state matrices, complete linkage clustering

Source
Second-generation PLINK: rising to the challenge of larger and richer datasets.
Embedded
yes

Text

The PLINK 1.0 –cluster –matrix flag combination launches an identity-by-state matrix calculation and writes the result to disk, and then performs complete linkage clustering on the data; when –ppc is added, a pairwise population concordance constraint is applied to the clustering process. As discussed earlier, PLINK 1.9 employs an XOR/bit population count algorithm which speeds up the matrix calculation by a large constant factor; the computational complexity of the clustering algorithm has also been reduced, from O(n3) to O(n2 logn). (Further improvement of clustering complexity, to O(n2), is possible in some cases [40].)