R-value thresholds influence module size and phylogenetic
relatedness of OTUs binned into a module
A key parameter in SCNIC is the R-value threshold used to pick modules.
Use of a high R-value threshold would be expected to bin only very
tightly correlated microbes with strong relationships, while less
stringent thresholds may identify community-level patterns representing
more loosely connected microbial pairs. To illustrate this concept, we
binned OTUs into modules using the SMD method at R-value thresholds
between 0.2 and 1.0 using the HIV dataset. As expected, at lower R-value
thresholds, more OTUs were binned into modules and lower numbers of
modules of smaller average size were formed as the threshold increased
(Figure 4A). To illustrate the effects of R-values thresholds on the
nature of the identified modules, we compare SCNIC outputs using R-value
thresholds of 0.2, 0.4, and 0.65. As shown in Figure 4, which visualizes
modules in Cytoscape using SCNIC output files, the R-value threshold
influences the size and connectivity of the network. We also illustrate
the effect of using different thresholds by examining the correlations
between OTUs that are included in the first module output by SCNIC,
which is the largest module (module-0 ) (Figure 5). All OTUs in
module-0 are positively correlated with each other, since SCNIC
only captures positive correlations.
Microbes co-occurring in the same environmental niche have previously
been observed to be phylogenetically closer on average[4]. This is
likely because phylogenetic relatedness has been correlated with
functional relatedness, such as through having more shared genome
content, leading towards success in similar environments [55]. We
show that increasing the R-value threshold results in modules that
contain OTUs that are more phylogenetically similar on average (Figure
5).