We discuss the ingress filtering strategy in Section 4.2.
Operations are carried out by the detection followed by filtering strategy.
By applying a three-stage filtering strategy on publicly available data from different sources, we identified 49 potential blood-detectable protein biomarkers.
We address filtering strategy, normalization methods, model comparison, and statistical power considerations below.
In contrast, the TC, UQ, DESeq, and RPKM normalization methods were robust to filtering strategy.
They also employed an energy threshold based filtering strategy to discard low energy samples.
We expect there will be no optimal statistical test, filtering strategy or frequency cut-off for gene-based tests.
Our filtering strategy leads to a list of 28 pathways involving 182 genes for further GxE investigation.
We recommend the DESeq or TMM normalization methods, noting that TMM is more sensitive to filtering strategy.
Next, we apply the widely used posterior filtering strategy of results obtained from genome wide studies to effectively reduce the amount of data obtained from individual platforms.
As Fig. 2 shows, the generalized linear model analysis using DESeq was the most sensitive to filtering strategy, while the ANOVA model was the least sensitive.
