Table 3 Running times of MST. Wall-clock time was measured in different parts of the pipeline: Time to build the input matrix, time to produce a new vector from an input FASTQ file and time to perform the MST of one sample. Except for the process named “Build source matrix”, the average time was estimated on the results from the validation set. MST done by FEAST does not allow for multithreading and was run using 2GB of memory and 1 core, whereas mSourceTracker cannot split one sink into multiple jobs, so 1 core and 250 GB of memory were allocated for each sink. Every other process was run using 250 GB of memory and 10 cores. Results for decOM are presented in bold

From: decOM: similarity-based microbial source tracking of ancient oral samples using k-mer-based methods