To ensure data quality, only the CpG dinucleotides that are covered by ≥5 bisulfite reads were retained (such CpG dinucleotides are designated as "sampled CpGs").
To ensure the accuracy of the methylome data, only CpG dinucleotides covered by ≥5 bisulfite/MethylC reads were considered (such CpG dinucleotides were designated as "sampled CpGs").
A total of 17 samples from male mice were analyzed, including 5 inorganic arsenic induced HCC samples (designated As-HCC), 5 samples of surrounding nontumorous, normal liver tissues from arsenic-treated mice (As-Normal), and 7 samples of age-matched control mouse liver (Control).
These three derived from patient 17, and the remaining 12 isolates, even those selected from cultures of the same sample, were nonetheless designated as L. iners.
The sample was assumed to be "additive free" and designated as purified sample.
If a terminal node contains the proportion of cancer sample (p) > 50% (i.e., p > 0.5), all the samples in this terminal are designated as cancer samples and p is the probability value assigned to the entire sample in this terminal node.
If a terminal node contains the proportion of tumor sample, (p) > 50% (i.e., p > 0.5), then all the samples in this terminal are designated as tumor samples, and p is the probability value assigned to the entire sample in this terminal node.
DF assigned a probability value for each prediction, where samples with the probability value ≥ 0.5 were designated as cancer samples, whereas others were designated as normal samples.
The samples that fell outside or near the periphery of the ellipse were designated as 'unique' samples for inclusion in the reference data set, as capturing this variance in a predictive model would most likely encompass future samples (Additional file 1: Figure S1).
In order to control this effect of overfitting, we split the data set of 334 students into two random subsamples A and B. Sample A, designated as the development sample, was used to construct various scales from the item pool.
