Thus, we were surprised to find that cRBH, a conservative clustering version of the simplest and earliest-developed of the four algorithms tested that drops instead of resolving inconsistencies [4], [6], [12], [13], [55], was consistently (e.g., across several parameter values and varying numbers of species) the best ortholog predictor.
Only clusters surviving at familywise error (FWE) or false discovery rate p <.05, corrected at cluster level with a conservative cluster-forming threshold, αc of.001, are reported.
The calculated sample size assumes a two-tailed probability level of 5%, a conservative clustering effect equivalent to an intraclass correlation coefficient of 0.10, and a regression model with 25 background covariates explaining 25% of the outcome variance (comparable to what might be expected from the selected variables).
The cluster in the upper right corner is a very conservative cluster dominated by the Tea Party.
Furthermore we assumed a conservative Intra Cluster Correlation of 0.10 for clustering of patients, based on the scarce literature on the correlation between long-term psychiatric patients within one professional [ 17].
Note, however, that "bostonbomber" is still one of the most frequently used words in conjunction with immigration in the most conservative cluster.
We adjusted the findings of the conservative cluster analysis by baseline differences.
After annotating each genome, a conservative gene clustering algorithm was used [39] to bin together orthologous genes from the different strains to ensure that simple allelic difference did not result in our calling orthologous loci as unique.
Cd-hit [ 31] was used to cluster reftigs with various sequence identity thresholds ranging from permissive clustering at 80% to conservative clustering at 99% using a k-mer word size of 5 to 10 increasing incrementally with the threshold (e.g. word size 5 with threshold 80%).
We used conservative clustering thresholds to ensure that each cluster contained relatively closely related sequences.
This is clearly an upper estimate, and given that the top 609 nucleotide clusters could be collapsed with less conservative clustering criteria into 85 larger clusters (see Materials and Methods), the actual number of transcripts could be 7-fold or more lower.
