Archives

  • 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • br J Biermann et al Genomics xxx

    2019-09-23


    J. Biermann, et al. Genomics xxx (xxxx) xxx–xxx
    A
    Stable genome
    Stable genome
    Unstable genome
    B
    Genome stability
    Stab
    Unstab
    C
    Dendrogram of gene BYL-719 data (n = 136)  E
    Fig. 1. Stratification of copy number and gene expression data by genome stability. (A) Frequency plots of copy number alterations distributed between tumours with stable genomes (G2I-1; n = 33), intermediately stable genomes (G2I-2; n = 61), and unstable genomes (G2I-3; n = 42). Hierarchical clustering was applied using Ward's method and Euclidean distance to identify underlying patterns in copy number (B) and gene expression data (D). Principal component analysis (PCA) of copy number (C) and gene expression data (E). Tumours with stable genomes (corresponding to G2I-1 and G2I-2) are coloured in blue and tumours with unstable genomes in pink. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
    transcripts were identified as significantly differentially expressed be-tween tumours with stable and unstable genomes. PCA based on these 335 transcripts showed a distinct separation between the genomically stable and unstable cluster (Fig. 2A). Applying t-distributed stochastic neighbour embedding (t-SNE) based on the 335 transcripts exhibited better discrimination between stable and unstable tumours (Fig. 2B). Interestingly, the few genomically stable tumours that clustered among the genomically unstable tumours belonged to the G2I-2 group (Sup-plementary Fig. S1A-B).
    Seventeen genes were selected based on the highest variable 
    importance in decision trees to define genome stability in a random forest model. PCA and t-SNE based on these 17 markers showed the ability to cluster the cohort according to genomic stability (Fig. 2C-D). The separation was less precise compared to the dimensional reduction incorporating 335 transcripts. Once more, the G2I-2 group formed the bridge between the unstable G2I-3 and the stable G2I-1 cluster (Sup-plementary Fig. S1C-D). The 335 differentially expressed genes segre-gated tumours with stable and unstable genome and showed an overlap with the risk groups based on the linear predictor of the 17-marker panel (Supplementary Fig. S2). The low-risk group contained 66% (62/
    J. Biermann, et al. Genomics xxx (xxxx) xxx–xxx
    Fig. 2. Dimensional reduction of differentially expressed genes. PCA (A) and t-SNE plots (B) of the 335 significantly differentially expressed genes identified by logistic regression. PCA (C) and t-SNE plots (D) for the 17 genes with the highest impact on genome stability identified by random forest modelling. Heatmap of the 17 markers included in the gene signature and overlap with genome stability, linear predictor-based risk groups, and ER, PR, and HER2 status (E). Tumours with stable genomes (corresponding to G2I-1 and G2I-2) are coloured in blue and tumours with unstable genomes in pink. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
    94) of tumours with stable genome, which consisted of 25 G2I-1 tu-mours and 37 G2I-2 tumours (Table 1; Supplementary Table S2; Sup-plementary Fig. S3). The high-risk group comprised 69% (29/42) of tumours with unstable genome along with 24 tumours that were clas-sified as G2I-2 and 8 tumours classified as G2I-1. The distributional differences were statistically significant for genomically stable and unstable tumours as well as for tumours classified as G2I-1, G2I-2 and 
    G2I-3 classified genome (Fisher's Exact Tests P < .001). Among the 17 markers, 16 were overexpressed in tumours with unstable genome; the only exception was seen in the PIEZO2 gene (Fig. 2E).