More Related Content
What's hot (19)
Similar to poster_Xingzhi2 (20)
poster_Xingzhi2
- 1. Poster template courtesy Faculty & Curriculum Support (FACS), Georgetown University School of Medicine Clonal assessment of func.onal muta.ons in cancer from whole-exome sequencing data In cancer, clonal evolution is characterized based on single nucleotide variants and copy number alterations. Previous methods accepted as input all variants predicted by variant-callers, regardless of differences in dispersion of variant allele frequencies (VAFs) due to uneven depth of coverage and possible presence of strand bias, prohibiting accurate inference of clonal architecture. We present a general framework for assignment of functional mutations to specific cancer clones, which is based on distinction between passenger variants with expected low dispersion of VAF versus putative functional variants, which may not be used for the reconstruction of cancer clonal architecture but can be assigned to inferred clones at the final stage. It takes into account VAFs and genotypes of corresponding regions together with information about normal cell. Abstract Student: Xingzhi Chen1; Faculty Advisor: Xiaogang (Simon) Zhong, PhD1,2 1Department of BiostaKsKcs, BioinformaKcs, and BiomathemaKcs, ; 2Lombardi Comprehensive Cancer Center; Georgetown University Medical Center, Washington, DC 20057 on a binomial distribution: where d is the depth of coverage of the variation, f is the number of reads supporting the variant and c is the sample contamination by normal cells. We then write the log likelihood function to maximize Where are weights of the possibility computed for a corresponding genotype By adding weights that for each variant sum to one, we favor mutations with the lowest number of copies. Each mutation is then attributed to its most likely possibility, which is the possibility with highest probability to belong to a clone. The number of clones is determined by minimization of the Bayesian Information Criterion (BIC). Priors can be provided by the user, randomly generated or determined by the k-medoids clustering on mutations. Fig1, Nine mutation clusters corresponding to inferred clones are shown in different colors. Results Discussions References Roth, A., Khattra, J., Yap, D., Wan, A., Laks, E., Biele, J., Ha, G., Aparicio, S., Bouchard-C?t¨¦, A., and Shah, S. P., et al. , 2014. PyClone: statistical inference of clonal population structure in cancer. Nature Methods , 11 (4):396_398. ?Jiao, W., Vembu, S., Deshwar, A. G., Stein, L., and Morris, Q., 2014. Inferring clonal evolution of tumors from single nucleotide somatic mutations. BMC Bioinformatics , 15 (1): 35. Khurana, E., Fu, Y., Colonna, V., Mu, X. J., Kang, H. M., Lappalainen, T., Sboner, A., Lochovsky, L., Chen, J., Harmanci, A., et al. , 2013. Integrative Annotation of Variants from 1092 Humans: Application to Cancer Genomics. Science (New York, N.Y.) , 342 (6154):1235587. In the framework of this study, we carried out Illumina paired-end sequencing for 22 neuroblastoma patients (tumor tissue at diagnosis, relapse and constitutional DNA). DNA material for each patient (lymphocytes, primary tumors and relapse tumors) was in each case sequenced using Illumina HiSeq 2500 instruments to an average depth of coverage of 80 by Beijing Genomics Institute (BGI). Sequenced reads were mapped to the human genome hg19 using BWA, reads then were realigned around indels with the Genome Analysis ToolKit, followed by a base recalibration, and mutations were called using Varscan2. We performs clustering of cellular prevalence values of mutations defined by , where is the cellular prevalence, the number of copies of the corresponding locus, NC the number of chromosomal copies bearing the mutation, and P the tumor purity. We use an EM algorithm based on the probability to observe a specific number of reads confirming a mutation given the number of reads overlapping the position, the contamination and the cellularity of a clone. In more detail, we attribute to each possibility a probability to observe f reads supporting the variant given that the latter belongs to a clone of cellular prevalence , based Methods Fig2, Illustration of the rule for assignment of mutations to (i) the ancestral clones (ii) clones expanding after the treatment and (iii) shrinking clones. Fig3, Absolute numbers and proportions of deleterious mutations in gene maps and modules in the ancestral clones, and clones expanding and shrinking at relapse. Here we propose a framework to detect functional mutations in cancer samples and associate the mutations to their corresponding clonal structure. It allows reconstruction of clonal populations based on both variant allele frequencies and genotype information. It is based on two central ideas. First, high-reliability passenger mutations must be used to reconstruct the clonal structure of tumors samples; then, low coverage functional mutations (with high variance in VAFs) should be mapped onto the inferred clonal structure. Second, we suggest to limit the set of functional mutations to those in genes known to be associated with cancer (e.g., Cancer Census genes) or to those in genes from gene modules/pathways frequently disrupted in a given cancer type. In this application, we excluded information about SVs and indels due to the low mapping. In the future, the sensitivity can be improved by using higher depth of coverage data and combining the paired-end datasets with reads produced with the mate-pair protocol or with long PacBio reads. ¦È = VAF ¡Á NCh NC ¡Á P NCh P( f |¦È) = d f ? ? ? ? ? ? ¦È ¡Á NC(1? c) NCh ? ?? ? ?? f ¡Á 1? (¦È ¡Á NC(1? c)) Nch ? ?? ? ?? d? f L = ¦Øi,p ti,k log Pi,s,p ( fi,s,p |¦Èk,s ) p ¡Æ s ¡Æ k ¡Æ i ¡Æ , ¦Øi,p ¦Øi,p = xs NCi,s,p ? ? ? ? ? ? ? ? + ys NCi,s,p ? ? ? ? ? ? ? ? 2 NChs s ¡Ç