Figure 1.
IBD mapping of founder mutations in recurrent regions of LOH.
This diagram illustrates the principles underlying our method. A founder mutation (red star on the schematic diagram of chromosomes) spreads through a population within a chromosome fragment (in red) inherited from the ancestral founder (A). Due to crossing-overs (dashed lines) between homologous chromosomes at meiosis, this chromosome fragment is shortened over generations, such that mutation carriers (indicated in red) eventually harbor only a short identical by descent (IBD) haplotype around the mutant gene (B). In addition, the wild-type counterpart of germline mutations is frequently lost by LOH in tumors, such that the founder mutation typically lies within a minimal region of LOH (C). As a result, the founder mutation is located within a haplotype conserved in each mutation carrier (peak IBD score), in the minimal region of LOH (D).
Figure 2.
Haplotype inference from tumor SNP profiles.
Genotypes can be inferred from SNP array data, with the B allele frequency (BAF), which characterizes the fluorescence ratio between the A and B alleles at each locus: BAF = B/(A+B). In constitutive DNA, each SNP is present as two alleles. The genotype of each SNP (AA, AB, or BB) can thus be determined from the BAF (0, 0.5 or 1, respectively), but not the haplotype of each chromosome. By contrast, if one of the two copies has been lost by LOH in the tumor, the tumor SNP profile directly reflects the haplotype of the chromosome that is retained in the tumor. If paired constitutive and tumor DNA samples are available, the haplotype of the lost chromosome can be reconstructed by comparing the two profiles.
Figure 3.
Detecting significantly recurrent IBD in a set of tumor haplotypes.
The FounderTracker pipeline for recurrent IBD detection is divided into two steps. An IBD score is first calculated for each SNP marker in the tumor set of haplotypes. The IBD score is derived from the pairwise IBD segments identified with the GERMLINE algorithm. It takes into account their length and the allelic frequencies of SNPs within each segment. The IBD score for tumors is then compared with a null distribution established from a reference data set, to identify the SNPs with IBD scores significantly higher than expected by chance.
Figure 4.
A conserved haplotype around the TP53 founder mutation in childhood adrenocortical tumors.
The IBD score calculated along chromosome 17 for the set of 13 childhood adrenocortical tumors (ACT) displays a significant peak at 17p13.1 (A). The sharp increases in the IBD score curve correspond to the boundaries of IBD segments in each pair of tumors. The peak region is represented in detail in panel B, with the haplotypes of the 13 chromosome copies retained in the tumors (harboring the mutant TP53 allele), and the haplotypes of the chromosome copies lost by LOH (harboring a wild-type TP53 allele) reconstructed for the six patients with matched normal blood samples. The haplotypes of wild-type alleles are all different, demonstrating the existence of different haplotypes for this genomic region in the population analyzed. By contrast, a haplotype common to all mutant alleles spans 470 kb upstream and 32 kb downstream from TP53 (represented by a thick red line). This conserved haplotype corresponds to the chromosome fragment harboring the R337H TP53 mutation inherited from the ancestral founder.
Figure 5.
Detection of a long conserved haplotype around SDHD gene in two paragangliomas.
(A) LOH analysis revealed 10 chromosome arms with LOH frequency >20% in our set of 30 pheochromocytomas and paragangliomas (top). These regions were analyzed using FounderTracker to detect conserved haplotypes, revealing a single significant region on chromosome arm 11q (bottom). (B) Visualization of the significant region identified on chromosome 11 with the Integrative Genomics Viewer [44]. The IBD score is represented as a blue line above tumor haplotypes. Haplotypes are represented as series of blue and yellow vertical lines, corresponding to SNPs with respectively “A” and “B” genotype, according to Illumina nomenclature. The significant region detected by FounderTracker corresponds to the long haplotype that is identical in tumors HS_048 and HS_158, and results in a high IBD score for this segment. This region contains 32 genes, including SDHD (indicated in red).
Table 1.
Power to detect conserved haplotypes as a function of haplotype length and prevalence.
Figure 6.
Power analysis with simulated data.
The performance of FounderTracker was assessed under various conditions, by comparison with the DASH method. The ability of each method to detect conserved haplotypes was established as a function of haplotype length (1 to 5 cM) and prevalence (2 to 10% of the samples). For each condition, the mean ROC curve was established by applying each method to 100 simulated datasets.