However, both of these methods are computationally intensive, and cannot run on even moderately large datasets (e.g., BEST is slower than *BEAST, and *BEAST is too computationally intensive to use on datasets with more than about 100 loci) [ 30, 29].
For example, MITIE takes 163.7X longer than ISP (and 92.4X longer than Cuffmerge) to run when 7 samples are used, making it difficult to use on datasets with a large number of samples, especially when the large transcriptomes such as human or mouse are being studied.
This indicates that the GCPV method does group together functionally related gene clusters effectively and can be used on larger datasets with gene clusters that are not well-characterized.
We evaluated the feasibility of using ASTRAL on datasets with large numbers of taxa using the 100-taxon simulated datasets, with 25 genes and 10 replicates.
The SOTA algorithm used on this dataset with linear correlation distance with 0 offset, 1000 cycles and 1.01 variability threshold parameters, led to the detection of 19 clusters.
The system was developed using one dataset and might not perform well on other datasets, especially with the rules that were developed.
We validated ParaSAM with samr using one-group, two-group and multigroup datasets using 1000 permutations on datasets with 3000 genes for 12 arrays (One class), 14 arrays (Two class) or 25 arrays (Multiclass).
SWang test can be used not only on datasets with large sample size but also on those with small sample size, the degrees of which depend on the sample size and the moment.
Moreover, the score is based on the original ratio of active compounds in the entire dataset and can hardly be used to compare predictions on datasets with different class distributions.
The old version of HMMerThread was used on the same dataset with standard settings and a hit-depth of 25.
