Figure 1

Transformation-based PCA of age groups using Pfams. Transformation-based PCA of age groups using the abundance of Pfams in the age-balanced dataset. Each color marks one age decade, except for red which corresponds to samples 70 to 90 years old. The first row used all Pfams present in each samples, the second and third rows used only top selected Pfams at age cutoff 40 by TF-iDF and mRMR separately. The columns are PC1 vs. PC2, PC1 vs. PC3 and PC2 vs. PC3 respectively. There is great overlap between most age groups, which explains why classification may be difficult. Besides the differences induced by source country of the metagenomes, mRMR reduces the variation within groups while TF-iDF facilitates linear classification. mRMR, minimum-redundancy maximum-relevancy; PCA, principal component analysis; Pfam, protein family; TF-iDF, term frequency-inverse document frequency.