Additional file 4: of Exome sequencing of primary breast cancers with paired metastatic lesions reveals metastasis-enriched mutations in the A-kinase anchoring protein family (AKAPs)

Figures S1a and S1b. Exome-wide view of mutations in metastatic breast cancer. Each panel (numbered by patient ID) shows whole-exome data from one patient in several aligned tracks, as indicated on the right. SNV: single-nucleotide variations, shown as tick marks for COSMIC recurrent cancer genes (outer tracks) and as density for all mutations (inner tracks). The central colored density gradient shows major allele concordance P value between primary and metastasis (50 SNP moving windows; blue: concordant; red: discordant), indicating the presence of concordant loss of heterozygosity. CNV: copy-number variation (red: deletion, blue: amplification); LOH: loss-of-heterozygosity. The Tumorscape track shows known hotspots for copy-number variation in breast cancer using data from Tumorscape Copy Number Alterations Across Multiple Cancer Types Release 1.6 (Broad Institute). CNV was determined by calculating the normalized read coverage at each SNP position for paired tumor and germline samples, then taking the log2 of the ratio of these numbers and recentering to zero. In Fig. 1, these measures were plotted as a moving average across ten adjacent SNPs. LOH: loss of heterozygousity) as major allele frequency for heterozygous SNPs on the vertical axis, with manually called LOH regions indicated by horizontal lines (vertical axis range 50–100%; red: significant SNP phase concordance with the paired sample, p < 0.01 by the binomial distribution; black: not significant (see also Additional file 3: Figure S2 a-b). Note that LOH could not be called for some samples (metastases of P1, P2, P4 and P10) because of low data quality. SNP phase concordance was determined using all SNPs in the LOH region in A and recorded the major allele of each (i.e. the ‘phase’ of the LOH region). We then tested the null hypothesis that the alleles of the corresponding SNPs in sample B was randomly distributed relative to A, with a 50% chance of concordance at each position. This would be expected if there were in fact no LOH in B, as the major allele would then be determined by the random fluctuations of read coverage. Under this model, the probability of observing k concordant calls among n SNPs is distributed according to the Binomial distribution (k; n, p) with p = 0.5. Thus we calculated the P value as the cumulative density of this distribution from k to n (that is, the probability of observing as many as k concordant calls, or more). A low P value (e.g. P < 0.01) on this test suggest that the LOH region in A was in fact also present on B, whereas a high P (e.g. P > 0.99) value suggests that there was LOH in B, but it was derived from the opposite allele compared with A. Intermediate values are expected whenever there was no LOH in sample B. The central track for each patient in Additional file 4: Figure S1 shows this P value. (ZIP 1670 kb)