Search

Additional file 6. Fig. S1. Kaplan-Meier plots of poor prognostic biomarkers. Fig. S2. Characteristics of the HRD-high cluster. Fig. S3. Characteristics of HRD-high tumors co-occurring with TP53 mutation. Fig. S4. Kaplan-Meier plots of expression-based prognosis markers. Fig. S5. Proliferative cluster enriched in poor prognostic markers. Fig. S6. Integrative analysis identified distinct prognostic clusters. Fig. S7. Molecular characteristics of integrative clusters. Fig. S8. Subtype switching driven by changes in PAM50 score composition. Fig. S9. Increased tumor growth and proliferation at PD. Fig. S10. IHC analysis of cell cycle markers. Fig. S11. Landscape of PD-specific genomic alterations. Fig. S12. RB1 LOF associated with increased APOBEC signature at PD. Fig. S13. APOBEC signature enriched in PD-specific tumor subclones.

Additional file 2: Figure S1. Mutation landscape of syngeneic models and types of somatic variants for each model. Figure S2. A. Ratio of Ts (Transition) to Tv (Transversion) substitution mutation in each syngeneic model; dashed lines represent maximum, median and minimum of median Ts/Tv from each human tumor types (data from (Alexandrov et al. 2013)) respectively. B. Single nucleotide mutation changes in each syngeneic model. Figure S3. E-cadherin and vimentin stain in syngeneic models. Figure S4. A. Correlation of neoantigen with mutational load in syngeneic models. B. Correlation of neoantigen with cytolytic activity in syngeneic models. The correlation was calculated using spearman method in R. numMutation: number of missense mutation. Figure S5. Response of syngeneic tumor models to anti-CTLA4, or anti PD-1. All mice were dosed intravenously. Individual tumor volumes are shown for 10 mice treated with PBS (black) or 10 mg/kg anti-CTLA4 (9H10) (red trace), or 10 mg/kg anti-PD-1 (RMP 1–14) (blue trace). All 10 mice bearing CT26 tumors dosed with anti-CTLA4 had no measurable tumor from study day 25 until measurements ended on study day 238. Comparison of mean tumor volumes was analyzed using log-transformed ANOVA. P-values are shown if there was a statistical difference between vehicle and antibody-treated groups. n.s. not significant. Figure S6. A. Expression of Gzma and Prf1 in syngeneic model grown in vitro (Cell) and in vivo (Tissue) across models based on protein expression. B. Cytolytic activity of syngeneic models. Cytolytic activity (CYT) is defined as the log-average (geometric mean) of Gzma and Prf1 protein expression. C. Protein expression of Gzma in soluble fraction. LFQ: label free quantitation. Figure S7. Gene expression of β-catenin, β-catenin target genes (gene list from (Spranger et al. 2015)) and epigenetic modulators (Ezh2, Dnmt1) across syngeneic models. Figure S8. CTLA4 expression in syngeneic in vivo tumor samples.

A schematic diagram of the GAP approach. Its matlab implementation is included as Additional file 2. (DOCX 189 kb)

Unsupervised hierarchical clustering of tumor cells by functional dependency. (A) Grouping of drop-out patterns using outlier genes, where rows correspond to genes and columns correspond to cell lines. Non-outlier genes are colored in grey as they represent non-sensitive hits. The right inset provides a zoom-in view of the yellow cluster with the vast majority of hematological lines. The full size heatmap, dendrograms and labels have been included as Additional file 6. (B) Grouping using outlier genes from the Wnt pathway. (DOCX 1164 kb)

The outlier pattern for APC. (A) The kernel density plot of ATARiS gene level score for APC and details on the five outlier cell lines most vulnerable to its knockdown. (B) Unsupervised hierarchical clustering of tumor cells by functional dependency on outlier genes from the Wnt pathway, with APC highlighted in yellow. (DOCX 395 kb)

An implementation of the GAP approach in matlab. (PDF 43 kb)

An implementation of the GAP approach in matlab. (PDF 43 kb)

A schematic diagram of the GAP approach. Its matlab implementation is included as Additional file 2. (DOCX 189 kb)

Unsupervised hierarchical clustering of tumor cells by functional dependency. (A) Grouping of drop-out patterns using outlier genes, where rows correspond to genes and columns correspond to cell lines. Non-outlier genes are colored in grey as they represent non-sensitive hits. The right inset provides a zoom-in view of the yellow cluster with the vast majority of hematological lines. The full size heatmap, dendrograms and labels have been included as Additional file 6. (B) Grouping using outlier genes from the Wnt pathway. (DOCX 1164 kb)

The outlier pattern for APC. (A) The kernel density plot of ATARiS gene level score for APC and details on the five outlier cell lines most vulnerable to its knockdown. (B) Unsupervised hierarchical clustering of tumor cells by functional dependency on outlier genes from the Wnt pathway, with APC highlighted in yellow. (DOCX 395 kb)

The ATARiS gene level score distribution for (A) EGFR and (B) ERBB2. A probability density estimate is computed by Gaussian kernel smoothing. (DOCX 49 kb)

The ATARiS gene level score distribution for (A) EGFR and (B) ERBB2. A probability density estimate is computed by Gaussian kernel smoothing. (DOCX 49 kb)