Skip to main content
Fig. 4 | Genome Biology

Fig. 4

From: Ultrasensitive deletion detection links mitochondrial DNA replication, disease, and aging

Fig. 4

Unsupervised exploratory analysis of mtDNA deletion patterns. Hierarchical clustering revealed similarities between samples and principal component analysis (PCA) revealed the underlying patterns that explain the most variation between samples. Both analyses used the fraction of deletions, binned by terminus positions (80 × 80 bins), rather than deletion frequencies or mtDNA ablation levels. Deletions of < 40 bp and deletions mapped to the T-shaped white region (b, c) were excluded as possible false positives (see Methods). a Hierarchical clustering by Euclidean distance separates samples into three clades. Gwt samples form a WT clade and sort largely by age at biopsy (black to red; see Additional file 1: Table S1). Gvar samples are “paraphyletic,” i.e., some have patterns that are more like Gwts (AT-WS clade) than like other Gvars (TI-PL clade). Colored squares below the dendrogram indicate POLG genotypes, with ablation levels shown below. Similar Gvar genotypes tend to cluster. Purple bars below the dendrogram illustrate ablation levels. b, c A representative subset of Gvar samples was used for PCA (see the “Materials and methods” section): all Gwts; AT-WS and TI-PL clade representatives; and cladeless sample M24. b The 1st PCA component vector (deletion terminus positions as per Fig. 3b) colored by PCA weight (red to blue for ≥ 0.04 to ≤ − 0.04). c The 3rd PCA component weight vector. dg Samples were scored using PCA weights. Black curves are parabolas fit to Gwt data (R2 inset). d The 1st component distinguishes Gwt (triangles) from Gvar deletion patterns, except M24 (orange diamond with black border). e 3rd component scores correlate with age at biopsy, regardless of POLG disease state. f These two components separate Gwt and Gvar clusters by deletion pattern (naïve to disease, age, and overall deletion frequency). g The 2nd PCA component was spent describing unique features of the disease state M24. In order to display minor arc features, the vectors in b, c were recalculated without M24

Back to article page