BATCH-SCAMPP: Scaling phylogenetic placement methods to place many sequences
Wedell, E.; Shen, C.; Warnow, T.
Show abstract
Phylogenetic placement, the problem of placing sequences into phylogenetic trees, has been limited either by the number of sequences placed in a single run or by the size of the placement tree. The most accurate scalable phylogenetic placement method with respect to the number of query sequences placed, EPA-ng, has a runtime that scales sublinearly to the number of query sequences. However, larger phylogenetic trees cause an increase in EPA-ngs memory usage, limiting the method to placement trees of up to 10,000 sequences. Our recently designed SCAMPP framework has been shown to scale EPA-ng to larger placement trees of up to 200,000 sequences by building a subtree for the placement of each query sequence. The approach of SCAMPP does not take advantage of EPA-ngs parallel efficiency since it only places a single query for each run of EPA-ng. Here we present BATCH-SCAMPP, a new technique that overcomes this barrier and enables EPA-ng and other phylogenetic placement methods to scale to ultra-large backbone trees and many query sequences. BATCH-SCAMPP is freely available at https://github.com/ewedell/BSCAMPP_code.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.