Quality scores and SNP detection in sequencing-by-synthesis systems

Promising new sequencing technologies, based on sequencing-by-synthesis (SBS), are starting to deliver large amounts of DNA sequence at very low cost. Polymorphism detection is a key application. We describe general methods for improved quality scores and accurate automated polymorphism detection, and apply them to data from the Roche (454) Genome Sequencer 20. We assess our methods using known-truth data sets, which is critical to the validity of the assessments. We developed informative, base-by-base error predictors for this sequencer and used a variant of the phred binning algorithm to combine them into a single empirically derived quality score. These quality scores are more useful than those produced by the system software: They both better predict actual error rates and identify many more high-quality bases. We developed a SNP detection method, with variants for low coverage, high coverage, and PCR amplicon applications, and evaluated it on known-truth data sets. We demonstrate good specificity in single reads, and excellent specificity (no false positives in 215 kb of genome) in high-coverage data.     

Comprehensive assessment of cancer missense mutation clustering in protein structures

Large-scale tumor sequencing projects enabled the identification of many new cancer gene candidates through computational approaches. Here, we describe a general method to detect cancer genes based on significant 3D clustering of mutations relative to the structure of the encoded protein products. The approach can also be used to search for proteins with an enrichment of mutations at binding interfaces with a protein, nucleic acid, or small molecule partner. We applied this approach to systematically analyze the PanCancer compendium of somatic mutations from 4,742 tumors relative to all known 3D structures of human proteins in the Protein Data Bank. We detected significant 3D clustering of missense mutations in several previously known oncoproteins including HRAS, EGFR, and PIK3CA. Although clustering of missense mutations is often regarded as a hallmark of oncoproteins, we observed that a number of tumor suppressors, including FBXW7, VHL, and STK11, also showed such clustering. Beside these known cases, we also identified significant 3D clustering of missense mutations in NUF2, which encodes a component of the kinetochore, that could affect chromosome segregation and lead to aneuploidy. Analysis of interaction interfaces revealed enrichment of mutations in the interfaces between FBXW7-CCNE1, HRAS-RASA1, CUL4B-CAND1, OGT-HCFC1, PPP2R1A-PPP2R5C/PPP2R2A, DICER1-Mg²⁺, MAX-DNA, SRSF2-RNA, and others. Together, our results indicate that systematic consideration of 3D structure can assist in the identification of cancer genes and in the understanding of the functional role of their mutations.To elucidate the genetic basis of cancer, efforts have been initiated to sequence the exomes or genomes of many human tumors. Among them are large-scale efforts such as The Cancer Genome Atlas (TCGA) (1) and the International Cancer Genome Consortium (ICGC) (2), as well as many smaller-scale projects. These efforts have collectively found millions of somatic mutations in virtually all human genes (the vast majority of which are nonfunctional or “passenger” mutations) across thousands of tumor samples (3). The amounts of available cancer sequencing data are growing rapidly and will continue to grow in the foreseeable future. We and others have developed computational methods to detect cancer-associated genes and functional mutations from such data, based on a significant overall burden of mutations or on significant positional clustering of mutations in the one-dimensional (1D) gene sequences, corresponding to mutational hotspots (3–6).For some cancer proteins, it has been observed that, although mutations may be distributed along the linear amino acid sequence, they tend to cluster in certain regions of the 3D structure, such as active sites. A clear example is KRAS, where particular missense mutations at the active site are positively selected in cancer because they disable the GTPase activity of the protein, locking it in its GTP-bound, active state, which promotes proliferation. As a result, recurrently mutated residues (e.g., G12, G13, I36, A59, Q61, K117, A146) tend to occur around the substrate-binding pocket of KRAS (Fig. 1). This and other individual examples of proteins showing 3D clustering of cancer missense mutations are sometimes used in the literature as supporting evidence for the involvement of those proteins in the disease or as a basis for functional hypotheses about the clustered mutations [e.g., EGFR (8), PIK3CA (9), DIS3 (10), SPOP (11), MRE11 (12), ERCC2 (13)]. Stehr et al. (14) and Ryslik et al. (15) assessed the structural clustering of missense mutations in 29 and 131 proteins, respectively, and demonstrated that taking into account 3D structural information can be helpful for identifying mutation hotspots in known cancer proteins or in new candidates.Open in a separate windowFig. 1.Spatial mutation clustering in KRAS. (A) Protein sequence of KRAS (Isoform 2B; UniProt: P01116-2) with mutated residues from the PanCancer data set (3) shown in red. Recurrent mutations (at least three samples) are shown in larger font and are annotated with position and number of samples with such mutations. Gray arcs between such residues are shown if their centroids are located closer than 13 Å between each other in the protein structure; arc width and label show the spatial distance in these cases (wider arcs corresponding to shorter distances). The C-terminal part of the protein sequence not covered by the structure in B is shown in smaller, gray font. (B) 3D structure of KRAS (gray) with substrate GDP (blue) bound to its active site (PDB ID code 4LUC) (7). Mutated residues are shown in red (recurrent mutations: sticks, nonrecurrent mutations: thin lines) and color intensity scales with the number of mutations per residue.Here, we seek to undertake comprehensive studies of 3D clustering of somatic missense mutations in cancer across all human proteins with available protein structures. Such integrative analysis may help to identify new cancer proteins that have been missed by other methods. In addition, it can help explain the functional roles of individual mutations based on their spatial location in the protein; for example, mutations that cluster at protein interaction interfaces may perturb key molecular interactions (16).     

Allelotyping of butadiene-induced lung and mammary adenocarcinomas of B6C3F1 mice: frequent losses of heterozygosity in regions homologous to human tumor-suppressor genes.

To identify the potential involvement of tumor-suppressor gene inactivation during neoplastic development in B6C3F1 mice, genetic losses were determined from allelotypes of butadiene-induced lung and mammary adenocarcinomas. By using length polymorphisms in restriction fragments and simple sequence repeats, or "microsatellites," markers on each autosome were analyzed for allele losses in tumor DNAs. Losses of heterozygosity on chromosome 11 were observed at several loci surrounding the p53 tumor-suppressor gene (Trp53) in 12 of 17 mammary tumors and 2 of 8 lung tumors. Although most of these alterations appeared to result from nondisjunction, at least two examples of somatic recombination or deletion were also observed. Southern analysis revealed a homozygous deletion of the remaining Trp53 allele of one of these mammary tumors. Losses of heterozygosity were also detected at the Rb-1 tumor-suppressor gene in 7 of 17 mammary tumors and 1 lung tumor. Finally, frequent allele losses were observed on chromosome 4 in lung tumors. Analysis of nine chromosome 4 loci defined an interstitial deletion containing the Ifa gene cluster in one of the lung tumors. A tumor-suppressor gene was previously mapped to this region of chromosome 4 in studies with somatic cell hybrids. In addition, homozygous deletions have been reported in a homologous region of human chromosome 9p for acute lymphocytic leukemias, glioblastomas, melanomas, and lung carcinomas. These findings suggest that the inactivation of tumor-suppressor genes including Trp53, Rb-1, and an unidentified gene on chromosome 4 plays a significant role during carcinogenesis in mice.     

Restriction fragment length polymorphism linkage map for Arabidopsis thaliana

We have constructed a restriction fragment length polymorphism linkage map for the nuclear genome of the flowering plant Arabidopsis thaliana. The map, containing 90 randomly distributed molecular markers, is physically very dense; greater than 50% of the genome is within 1.9 centimorgans, or approximately 270 kilobase pairs, of the mapped DNA fragments. The map was based on the meiotic segregation of markers in two different crosses. The restriction fragment length polymorphism linkage groups were integrated with the five classically mapped linkage groups by virtue of mapped mutations included in these crosses. Markers consist of both cloned Arabidopsis genes and random low-copy-number genomic DNA clones that are able to detect polymorphisms with the restriction enzymes EcoRI, Bgl II, and/or Xba I. These cloned markers can serve as starting points for chromosome walking, allowing for the isolation of Arabidopsis genes of known map location. The restriction fragment length polymorphism map also can associate clones of unknown gene function with mutant phenotypes, and vice versa.