We are searching data for your request:
Upon completion, a link will appear to access the found materials.
I am looking for a statistical approach to inter-codon mutations. For example a 64*64 (64*63 actually) table, that contain the possibility of mutation from one codon to another codon (CCA to CAA or CGG… for example).
Is there any articles, databases or others that provide such table?
Please comment if my question is not completely clear. I searched but didn't find an answer for this question in Google Scholar.
MUTATION DETECTION AND SCREENING
Mutations are genetic alterations that are acquired in germ or non-germ (somatic) cells. Mutations can be present as an insertion, deletion, or base pair change in the coding or non-coding regions, resulting in silent, missense, or nonsense mutations. In some cases, a mutation occurs at the intron-exon boundary, disrupting the normal splicing of the transcript. Sanger sequencing-based mutation analysis, mutation screening, and exon resequencing all involve high volume PCR amplification and sequencing to uncover these mutations.
Mutation analysis and screening techniques can be used as either your primary source for mutation detection, or as a confirmation of next generation sequencing and microarray results. No matter the application, utilize GENEWIZ’s expertise in targeting genomic regions of DNA with specific, robust assays.
Statistical coupling analysis (SCA) is a method for analyzing multiple sequence alignments that was used to identify groups of coevolving residues termed “sectors”. The method applies spectral analysis to a matrix obtained by combining correlation information with sequence conservation. It has been asserted that the protein sectors identified by SCA are functionally significant, with different sectors controlling different biochemical properties of the protein. Here we reconsider the available experimental data and note that it involves almost exclusively proteins with a single sector. We show that in this case sequence conservation is the dominating factor in SCA, and can alone be used to make statistically equivalent functional predictions. Therefore, we suggest shifting the experimental focus to proteins for which SCA identifies several sectors. Correlations in protein alignments, which have been shown to be informative in a number of independent studies, would then be less dominated by sequence conservation.
Overview of the study design
Our analysis included 33 algorithms (reported in 29 studies) that could prioritize or categorize SNV mutations that result in amino acid changes. To robustly assess the performance of different algorithms, we employed five different benchmark datasets: (i) the mutation clustering patterns in protein 3D structures (ii) literature annotation based on OncoKB , a widely used knowledge database in the cancer research community (iii) the effects of TP53 mutations on their target transcription activity (iv) the effects of cancer mutations on tumor formation in xenograft experiments and (iv) functional annotation based on in vitro cell viability assays developed by our group. These benchmark datasets represent different features of driver mutations relative to passenger mutations and are highly complementary to each other, thereby ensuring a comprehensive assessment. Given the positive (driver) and negative (passenger) cases defined in each benchmark dataset, based on numeric scores for each algorithm, we employed area under the curve (AUC) of receiver operating characteristics (ROC) curves to assess the predictive performance, which is a common measurement independent from the threshold value in each algorithm. In addition, we compared categorical predictions of different algorithms against true labels in each benchmark analysis (Table 1, Additional file 1).
Table 1 shows the characters of the 33 algorithms we assessed in this study. Among them, six algorithms were developed specifically to predict cancer driver mutations, and the others were designed to predict the functional impact of an SNV in general. While not developed for identifying cancer drivers, those non-cancer-specific algorithms, such as SIFT and Polyphen2, have been widely used to prioritize mutations in cancer-related research. Further, 16 are ensemble algorithms that use the scores from other published algorithms as input (Fig. 1a). These algorithms employ a variety of information as features to build predictive models: 10 use the features related to sequence context such as nucleotide change types and CpG island locations 9 contain protein features such as domain and amino acid changes 24 consider evolutionary conservation, and 6 include epigenomic information (Fig. 1a). To study the correlations of different algorithms, we compiled and calculated the scores of the 33 algorithms for
710,000 unique mutations detected in the TCGA whole-exome sequencing project across 33 cancer types by the Multi-Center Mutation-Calling in Multiple Cancers (MC3), [12, 35]. We then quantified their score similarities using Spearman rank correlations across all these mutations and found that the algorithm scores showed overall positive correlations (Fig. 1b). In the dissimilarity-based tree (Fig. 1b), the algorithms derived from the same study were always clustered together such as Eigen-PC and Eigen , SIFT4G  and SIFT , and MetaLR and MetaSVM , which is expected given they were built in a similar way.
Feature summary and inter-correlations between algorithms. a Based on features included, each algorithm was labeled as using ensemble score, sequence context, protein feature, conservation, or epigenomic information. The algorithms trained on cancer diver data or proposed to identify cancer drivers are labeled as cancer-specific. b Left: hierarchical clustering pattern of 33 algorithms based on
710,000 TCGA somatic mutations right, a triangle heatmap displays the Spearman rank correlation coefficient between any two algorithms
Benchmark 1: Mutation clustering patterns in the protein 3D structures
The functional impact of a specific mutation largely depends on its location in the protein 3D structure. Functional or driver mutations tend to form spatial hotspot clusters. In recent years, several computational algorithms have been developed to detect mutation clusters in the protein 3D space, which are able to detect rare mutations with validated functional impacts. From this perspective, we constructed a benchmark dataset based on the mutation 3D clustering patterns. We employed four spatial cluster algorithms (HotMAPs , 3DHotSpots , HotSpot3D , and e-Driver3D ) to predict putative mutation hotspots. We defined the consensus score as the number of the four tools that predicted each mutation to be within a 3D cluster (Fig. 2a). We found a strong enrichment of mutations with a high consensus score in known cancer genes (i.e., cancer gene census [CGC]) (p < 2.2 × 10 −16 , Fisher’s exact test see the “Methods” section Additional file 2).
Assessment using a benchmark dataset based on mutation 3D clustering pattern. a Overview of the assessment process. We used four computational algorithms to detect whether mutations are located within the protein 3D structural hotspots, each algorithm with one vote. The number of votes was defined as the consensus cluster score. A mutation with a score of ≥ 2 and in a cancer gene (i.e., cancer gene consensus) was considered as a positive case, and a mutation with a score of 0 and in a non-cancer gene was considered as a negative case. b ROC curves and corresponding AUC scores for the top 10 algorithms. c Boxplots showing the differences of AUC between two groups of algorithms with or without certain features. p value is based on the Wilcoxon rank sum test. d Sensitivity and specificity of each algorithm calculated by using the median score value as the threshold to make binary predictions. Error bars, mean ± 2SD
To compile the benchmark set, from the
710k TCGA mutations, we designated mutations with a high consensus score (≥ 2) in a known cancer gene as driver candidates (positive cases, n = 1429) and randomly selected the same number of mutations with a consensus score of 0 in non-cancer genes as passenger candidates (negative cases, n = 1429). We then evaluated the performance of the 33 algorithms using ROC curves. We found that the performance of different algorithms varied greatly, and the AUC score ranged from 0.64 to 0.97, with a median value of 0.79 (Fig. 2b Additional file 3). Six algorithms had a AUC score of > 0.9, including CTAT-cancer , CanDrA , CHASM , DEOGEN2 , FATHMM-cancer , and MVP . To confirm our results, we generated another same-size negative set of CGC mutations with a consensus score of 0, repeated the evaluation, and found a strong correlation of AUCs between the two evaluations (Pearson correlation, r = 0.97 Additional file 4). In terms of group-based comparison (Fig. 2c), cancer-specific algorithms performed much better than general algorithms (mean AUC 92.2% vs. 79.0%, Wilcoxon rank sum test, p = 1.6 × 10 −4 ), and ensemble scores showed higher AUC scores than others (mean AUC 84.3% vs. 78.7%, Wilcoxon rank sum test, p = 0.015).
To evaluate the performance of binary predictions, we calculated accuracy, sensitivity, specificity, PPV, and NPV (see the “Methods” section Additional file 5). In the analysis, we randomly selected 1000 positives and 1000 negatives to construct the benchmark sets and used the median score value of each algorithm as the threshold to make binary predictions. The process was repeated for 100 times to estimate mean and standard deviation for each metric. CanDrA showed the highest overall accuracy (mean = 0.91), followed by CTAT-cancer, CHASM, DEOGEN2, and FATHMM-cancer. The sensitivity and specificity for CanDrA, CTAT-cancer, CHASM, DEOGEN2, and FATHMM-cancer consistently ranked among the top ones (Fig. 2d). Some algorithms, such as MutationTaster2 , Integrated_fitCons , GenoCanyon , and LRT , had very unbalanced sensitivities and specificities. In addition, we calculated the same metrics for the 17 algorithms with the default categorical predictions (see the “Methods” section Additional file 6). CanDrA and DEOGEN2 showed the highest accuracy. The results in this section provide an overview of how well the algorithms distinguish mutations clustered in the 3D space from the isolated ones in the protein structures.
Benchmark 2: Literature-based annotation
Functional effects of specific mutations have been a major theme in cancer research over decades. Therefore, literature is a rich resource to define the role of somatic mutations in cancer development. OncoKB is a widely used, expert-guided, precision oncology knowledge base where the functional effects of somatic mutations in > 400 cancer-associated genes have been classified into four categories (oncogenic, likely oncogenic, likely neutral, and inconclusive) based on their biological and oncogenic effects and the prognostic and predictive significance reported in the literature .
Based on OncoKB annotation, we performed two comparisons for the algorithm evaluation: (i) oncogenic (positive cases) vs. likely neutral (negative cases) (773 vs. 497) and (ii) oncogenic + likely oncogenic (positive cases) vs. likely neutral (negative cases) (2327 vs. 497) (Fig. 3a). The two comparisons yielded highly consistent results in terms of the AUC scores (Pearson correlation r = 0.90 Fig. 3b). The likely oncogenic mutations reduced the overall AUC scores, probably due to inconsistent literature annotations for those mutations. The top 10 algorithms in the first comparison had very close AUCs, ranging from 0.71 to 0.75 (Fig. 3b Additional file 7). We did not observe significant differences for group-based comparisons (Additional file 8). For binary predictions, we calculated accuracy, sensitivity, specificity, PPV, and NPV (Additional file 9), by using randomly selected 400 positives and 400 negatives (see the “Methods” section). PROVEAN , VEST4 , and MPC  had the highest accuracy values (0.69, 0.69, and 0.68 respectively PROVEAN, VEST4, MPC, REVEL , FATHMM-cancer, CTAT-population  were the top ones in both sensitivity and specificity (Fig. 3c). In addition, we calculated the same metrics for the 17 algorithms with the default categorical predictions (see the “Methods” section Additional file 10). DEOGEN2 showed the best accuracy (mean = 0.70). These results provide insights into how well the algorithms predict driver mutations based on literature-driven evidence.
Assessment using a benchmark dataset based on OncoKB annotation. a Overview of the assessment process. The OncoKB database classifies mutations into four categories: oncogenic, likely oncogenic, likely neutral, and inconclusive. We considered “likely neutral” as negative cases, and we considered “oncogenic” mutations only or both “oncogenic” and “likely oncogenic” mutations as positive cases. b Bar plots showing the AUC scores of the 33 algorithms in the two comparisons. The red color is for oncogenic plus likely oncogenic vs. likely neutral, and green is for oncogenic vs. likely neutral. c Sensitivity and specificity of 33 algorithms. Error bars, mean ± 2SD
Benchmark 3: Effects of TP53 mutations on target-gene transactivation
TP53 is the most frequently mutated gene in human cancers, and the IARC TP53 database compiles various types of information on TP53 gene variants . The TP53 mutants had been functionally assessed based on the median transactivation levels, measured as percentage of wild-type activity, of 8 TP53 targets (WAF1, MDM2, BAX, h1433s, AIP1, GADD45, NOXA, and P53R2). We constructed a benchmark dataset by selecting TP53 mutations with transactivation level ≤ 50% as positive cases, and all others as negative cases.
The top five algorithms, ordered by AUC scores, were CHASM, CTAT-cancer, CTAT-population, DEOGEN2, and VEST4 (Fig. 4b Additional file 11). While a few algorithms had an AUC of
50%, the majority of the 33 algorithms were above 80% (Additional file 11). It should be noted that CanDrA, FATHMM-cancer, and FATHMM-disease appear to be gene-specific, as all TP53 mutations were predicted to be drivers. We suspect that these tools intrinsically give very high scores for mutations in well-known cancer genes. In terms of group-based comparisons (Additional file 12), algorithms that used epigenomic information had significantly lower AUCs than others (Wilcoxon rank sum test, p = 0.02) cancer-specific algorithms showed marginally significant than the other algorithms (Wilcoxon rank sum test, p = 0.08). We calculated the accuracies using median scores as the threshold to make binary predictions for each algorithm and found that their performance varied considerably among algorithms. CHASM was the most accurate one (mean AUC = 0.88) followed by CTAT-cancer and CTAT-population (Additional file 13). MetaSVM had the lowest accuracy (mean = 0.44). Several algorithms, including Integrated_fitCons, LRT, and SIFT, showed very unbalanced ranks of sensitivity and specificity (Fig. 4c), due to the fact that these algorithms provide the same scores for most mutations in this benchmark dataset. CHASM, CTAT-cancer, CTAT-population, VEST4, and DEOGEN2 had both good sensitivities and specificities. For the 15 algorithms that were provided with recommended cutoffs in their original studies, we calculated the same five performance metrics based on their explicit cutoffs (see the “Methods” section Additional file 14). These results present an informative view of how well the algorithms distinguish putative TP53 mutation drivers that had a high impact on target transcription activity from passengers.
Assessment using a benchmark dataset based on the transactivation effects of TP53 mutations. a Overview of the assessment process. Promoter-specific transcriptional activity was measured for 8 targets of p53 protein. Mutations with the median transcription activity ≤ 50% were used as positive cases, and others were used as negative cases. b ROC plot and AUC scores for the top 10 algorithms. c Sensitivity and specificity of 33 algorithms. Error bars, mean ± 2SD
Benchmark 4: In vivo tumor formation assays
A recent study employed an in vivo tumor formation assay to systematically assess the oncogenicity of a large number of mutant alleles curated from > 5000 tumors . In the assay, HA1E-M cell lines that stably expressed individual mutant allele were injected into mice. Mutant alleles that formed any tumor > 500 mm 3 by 130 days were considered as oncogenic mutations and thus used as positive cases in our study, and all other alleles were used as negative cases (Fig. 5a). Based on the functional annotation of such 71 mutations (45 positives vs. 26 negatives), we evaluated the 33 algorithms. Five algorithms, including CHASM, PROVEAN, PrimateAI , and REVEL, had an AUC score of > 70% (Fig. 5b Additional file 15), while six algorithms were < 60%. Cancer-specific algorithms did not outperform others (Additional file 16), and there were no significant differences for other group-based comparisons as well.
Assessment using a benchmark dataset based on in vivo tumor formation. a Overview of the assessment process. Cell lines stabling expressing mutant alleles were injected into mice. Mutations that could form any tumors greater than 500 mm 3 by 130 days were considered as functional mutations and used as positives, and other mutations were used as negatives. b ROC plot and AUC scores for the top 10 algorithms. c Sensitivity and specificity of 33 algorithms. Error bars, mean ± 2SD
Using the median scores as thresholds, we compared categorical predictions against the true labels. PROVEAN had the highest accuracy (0.72), followed by PrimateAI and CHASM (Additional file 17). Most algorithms had balanced rankings in sensitivity and specificity (Fig. 5c). However, MutationTaster2, GenoCanyon, and LRT were the top three in sensitivity, but had the lowest specificities. This is because these three algorithms gave the same scores for most mutations in this benchmark analysis. Categorical outputs, directly provided by 17 algorithms as outputs, showed PROVEAN the highest accuracy (mean accuracy = 0.71 Additional file 18). The results in this section provided insights into how those algorithms were able to differentiate cancer mutations with tumor formation potential from those that unlikely drive tumor formation.
Benchmark 5: In vitro cell viability assays
A common functional consequence of a driver mutation is to confer a preferential growth or survival advantage to the cell, and this effect can be directly assessed by cellular assays. We recently developed a systems-biology approach to test the functional effects of mutations on an individual basis using an in vitro system . Briefly, we generated bar-coded expression mutated open reading frame (ORF) clones by a HiTMMoB approach , and then tested the effects of mutated ORFs in IL-3-dependent Ba/F3 cells (a sensitive leukemia cell line, frequently used in drug screening) and EGF- and insulin-dependent MCF10A cells (a non-tumorigenic breast epithelial cell line) in parallel using a lentiviral approach, with wild-type counterparts as well as negative and positive experimental controls. Based on the effects on cell viability in the two cell models, we generated a consensus functional annotation for each tested mutation based on an “OR gate” logic. Mutations with detectable effects (i.e., activating, inactivating, inhibitory, and non-inhibitory) are considered as driver candidates (positive cases), whereas those without a notable effect (i.e., neutral) are considered as passengers. Using this approach, our recent study  reported the functional annotation of a large number of somatic mutations. To increase the robustness of our evaluation, we selected another
200 mutations from the TCGA mutation pool, performed the same cell viability assays, and obtained the informative functional annotations of 164 mutations (Additional file 19). We performed the algorithm assessment using three experiment-annotated datasets: (i) the published dataset (797 in total positive vs. negative: 321 vs. 476), (ii) the new dataset (164 in total positive vs. negative: 55 vs. 109), and (iii) the combined dataset (961 in total positive vs. negative: 376 vs. 585) (Fig. 6a Additional file 19).
Assessment using a benchmark dataset based on in vitro cell viability. a Overview of the assessment process. For each mutation, we performed cell viability assays in two “informer” cell lines, Ba/F3 and MCF10A. Consensus calls were inferred by integrating the functional effects observed in Ba/F3 and MCF10A. We considered activating, inactivating, inhibitory, and non-inhibitory mutations as positive cases, while neutral mutations were considered negative. b The ROC curves of the 33 algorithms based on a combined set of published mutations (Ng et al. ) and newly generated mutations in this study. c Bar plots showing the AUC scores of the 33 algorithms in the three datasets: new functional data (red), published functional data (green), and the combined set (blue). d Boxplots showing the differences of AUC between two groups of algorithms with or without certain features. p values are based on the Wilcoxon rank sum test. d Sensitivity and specificity of 33 algorithms. Error bars, mean ± 2SD
We found that the predictive power of different algorithms varied greatly. Based on the published dataset, the top three algorithms were CTAT-cancer (AUC = 77.0%), CHASM (AUC = 75.4%), and CanDrA (AUC = 72.9%) (Fig. 6b Additional file 20A). Based on the new dataset, the top three algorithms were PrimateAI (AUC = 81.4%), REVEL (AUC = 77.6%), and CTAT-cancer (AUC = 77.5%) (Fig. 6b Additional file 20B). Based on the combined dataset, the top algorithms were CTAT-cancer (AUC = 77.1%), CHASM (AUC = 75.7%), and PrimateAI (AUC = 74.0%), whereas a few algorithms had an AUC score close to 0.5 (Fig. 6b Additional file 20C). The new dataset generally resulted in higher AUC scores than the published dataset, with the largest differences observed for FATHMM-disease , MetaLR, and MetaSVM (AUC difference = 0.21, 0.14, and 0.14 respectively). These differences may be due to the intrinsic features of the benchmark mutation sets.
We used the combined dataset for downstream analyses. In group-based comparisons, cancer-specific algorithms were significantly better than the others (mean AUC 72.0% vs. 63.5%, Wilcoxon rank sum test, p = 7 × 10 −4 ). The top three algorithms by the overall accuracy were CTAT-cancer (mean = 0.70), PrimateAI (mean = 0.70), and CHASM (mean = 0.69) (Additional file 21). All the three algorithms were among the top ones in terms of sensitivity and specificity (Fig. 6d). For the 17 algorithms with default categorical predictions, we calculated the same metrics using the same benchmark set (Additional file 22). The top three algorithms were PrimateAI, PROVEAN, and DEOGEN2. As these experimental data (especially the new data) were generated independently from the algorithm development, these results provide a valuable assessment of how well the algorithms identify driver mutations with an effect on cell viability in vitro.
From the above sections, we evaluated the performance of different algorithms using five different criteria. Each benchmark uses an independent information source to define driver and passenger mutation candidates. The positive cases and the negative cases included in each benchmark dataset are quite distinct. For the positive cases, 3D clustering pattern, OncoKB annotation, transactivation of TP53 mutations, in vivo tumor formation assays, and in vitro cell viability assays contained 56.1%, 68.1%, 46.4%, 15.6%, and 54.5% unique mutations, respectively (Fig. 7a). The percentages of unique negatives were even higher (Fig. 7b).
Overall evaluation. a, b The overlapping summary of positive (a) and negative cases (b) in the five benchmark datasets. c Correlations of the performance ranks of the 33 algorithms based on the five benchmark datasets. d A heatmap showing the rank of the 33 algorithms based on each benchmark dataset. Ranks are labeled for the top five algorithms only. Red, higher ranks, and white, lower ranks. The features of the 33 algorithms are shown on the top, indicated by color (gray, no and black, yes)
The five benchmark analyses showed an overall good consistency: the highest Spearman correlation of AUC scores was observed between in vitro cell viability assay and 3D clustering patterns (Fig. 7c). Interestingly, despite the diversity of the benchmark data used, we observed a great convergence on a few top-performing algorithms (Fig. 7d, the top five algorithms highlighted for each benchmark). CHASM and CTAT-cancer ranked among the top 5 for four times, but they were not among the top in the OncoKB benchmark and DEOGEN2 and PrimateAI were among the top 5 for three times including OncoKB. A few others, including VEST4, PROVEAN, MPC, CanDrA, REVEL, CATA-population, and FATHMM-cancer, ranked among the top 5 in one or two benchmarks. Except for CTAT-cancer and REVEL which were solely based on published predictors, the top-performing algorithms employ a wide range of features, including published scores, sequence context, protein features, and conservation. Collectively, CHASM, CTAT-cancer, DEOGEN2, and PrimateAI may represent the best choice for predicting cancer driver mutations.
Population level analysis of evolved mutations underlying improvements in plant hemicellulose and cellulose fermentation by Clostridium phytofermentans
Background: The complexity of plant cell walls creates many challenges for microbial decomposition. Clostridium phytofermentans, an anaerobic bacterium isolated from forest soil, directly breaks down and utilizes many plant cell wall carbohydrates. The objective of this research is to understand constraints on rates of plant decomposition by Clostridium phytofermentans and identify molecular mechanisms that may overcome these limitations.
Results: Experimental evolution via repeated serial transfers during exponential growth was used to select for C. phytofermentans genotypes that grow more rapidly on cellobiose, cellulose and xylan. To identify the underlying mutations an average of 13,600,000 paired-end reads were generated per population resulting in ∼300 fold coverage of each site in the genome. Mutations with allele frequencies of 5% or greater could be identified with statistical confidence. Many mutations are in carbohydrate-related genes including the promoter regions of glycoside hydrolases and amino acid substitutions in ABC transport proteins involved in carbohydrate uptake, signal transduction sensors that detect specific carbohydrates, proteins that affect the export of extracellular enzymes, and regulators of unknown specificity. Structural modeling of the ABC transporter complex proteins suggests that mutations in these genes may alter the recognition of carbohydrates by substrate-binding proteins and communication between the intercellular face of the transmembrane and the ATPase binding proteins.
Conclusions: Experimental evolution was effective in identifying molecular constraints on the rate of hemicellulose and cellulose fermentation and selected for putative gain of function mutations that do not typically appear in traditional molecular genetic screens. The results reveal new strategies for evolving and engineering microorganisms for faster growth on plant carbohydrates.
Conflict of interest statement
Competing Interests: Qteros had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Funding from Qteros did not alter the authors’ adherence to all the PLOS ONE policies on sharing data and materials.
Figure 1. Schematic representation of the adaptive…
Figure 1. Schematic representation of the adaptive evolution process starting from an isogenic founder.
Figure 2. Growth, cellobiose utilization and ethanol…
Figure 2. Growth, cellobiose utilization and ethanol production of cellobiose adapted populations and the founder.
Figure 3. Growth and ethanol production of…
Figure 3. Growth and ethanol production of xylan-adapted populations and the founder.
Figure 4. Major fermentation product formation by…
Figure 4. Major fermentation product formation by cellulose adapted populations and founder after 10 days…
Figure 5. Genes and intergenic regions where…
Figure 5. Genes and intergenic regions where multiple mutations were detected.
Mutation hotspots which were…
Figure 6. Homology models suggest that the…
Figure 6. Homology models suggest that the selected mutations in an ABC transporter binding protein…
Figure 7. Homology modeling suggests that a…
Figure 7. Homology modeling suggests that a selected mutation in an ABC transporter transmembrane domain…
Figure 8. Localization of SNPs in Cphy…
Figure 8. Localization of SNPs in Cphy 3212 cellulose adapted lines.
Figure 9. Overview of carbohydrate sensing, saccharification…
Figure 9. Overview of carbohydrate sensing, saccharification and transport systems with the approximate location of…
We provide a comprehensive genome-wide analysis of somatic mutagenesis in human cells. Our model of basal mutagenesis offers an enhanced understanding of the unavoidable loss of genome integrity and the protective forces that counteract this process, including the stem-cell niche and DNA repair. The finding of cell-type-specific mutagen exposures and consequences on cell fate in the kidney are a proof of principle supporting the importance of understanding mutational processes active in healthy human cells to understand cancer. WGS data from single genomes constitute a precious tool for achieving the goal because they allow the analysis of the non-coding portion of the genome. Overall, our comprehensive classification of mutagenic processes introduces a novel perspective for clinical advancements in preventing cancer- and age-related diseases.
Since the very first sequences of the insulin protein were characterized by Fred Sanger in 1951, biologists have been trying to use this knowledge to understand the function of molecules.   He and his colleagues' discoveries contributed to the successful sequencing of the first DNA-based genome.  The method used in this study, which is called the “Sanger method” or Sanger sequencing, was a milestone in sequencing long strand molecules such as DNA. This method was eventually used in the human genome project.  According to Michael Levitt, sequence analysis was born in the period from 1969–1977.  In 1969 the analysis of sequences of transfer RNAs was used to infer residue interactions from correlated changes in the nucleotide sequences, giving rise to a model of the tRNA secondary structure.  In 1970, Saul B. Needleman and Christian D. Wunsch published the first computer algorithm for aligning two sequences.  Over this time, developments in obtaining nucleotide sequence improved greatly, leading to the publication of the first complete genome of a bacteriophage in 1977.  Robert Holley and his team in Cornell University were believed to be the first to sequence an RNA molecule. 
This work was supported by NIH grants 3R01MH101814-02S1, HHSN26820100029C, and 5U01HG006569. We would like to thank the Geuvadis Consortium, the GTEx Consortium, the members of the Lappalainen lab, the former GSA group at the Broad, and the bioinformatics team of the New York Genome Center. The Genotype-Tissue Expression (GTEx) Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health (commonfund.nih.gov/GTEx). Additional funds were provided by the NCI, NHGRI, NHLBI, NIDA, NIMH, and NINDS. Donors were enrolled at Biospecimen Source Sites funded by NCISAIC-Frederick, Inc. (SAIC-F) subcontracts to the National Disease Research Interchange (10XS170), Roswell Park Cancer Institute (10XS171), and Science Care, Inc. (X10S172). The Laboratory, Data Analysis, and Coordinating Center (LDACC) was funded through a contract (HHSN268201000029C) to The Broad Institute, Inc. Biorepository operations were funded through an SAIC-F subcontract to Van Andel Institute (10ST1035). Additional data repository and project management were provided by SAIC-F (HHSN261200800001E). The Brain Bank was supported by a supplement to University of Miami grant DA006227. Statistical Methods development grants were made to the University of Geneva (MH090941), the University of Chicago (MH090951 and MH090937), the University of North Carolina - Chapel Hill (MH090936) and to Harvard University (MH090948).
Molecular dynamics and mutational analysis of a channelopathy mutation in the IIS6 helix of Ca V 1.2
A channelopathy mutation in segment IIS6 of Ca(V)1.4 (I745T) has been shown to cause severe visual impairment by shifting the activation and inactivation curves to more hyperpolarized voltages and slowing activation and inactivation kinetics. A similar gating phenotype is caused by the corresponding mutation, I781T, in Ca(V)1.2 (midpoint of activation curve (V(0.5)) shifted to -37.7 +/- 1.2 mV). We show here that wild-type gating can partially be restored by a helix stabilizing rescue mutation N785A. V(0.5) of I781T/N785A (V(0.5) = -21.5 +/- 0.6 mV) was shifted back towards wild-type (V(0.5) = -9.9 +/- 1.1 mV). Homology models developed in our group (see accompanying article for details) were used to perform Molecular Dynamics-simulations (MD-simulations) on wild-type and mutant channels. Systematic changes in segment IIIS6 (M1187-F1194) and in helix IIS6 (N785-L786) were studied. The simulated structural changes in S6 segments of I781T/N785A were less pronounced than in I781T. A delicate balance between helix flexibility and stability enabling the formation of hydrophobic seals at the inner channel mouth appears to be important for wild-type Ca(V)1.2 gating. Our study illustrates that effects of mutations in the lower part of IIS6 may not be localized to the residue or even segment being mutated, but may affect conformations of interacting segments.
Structural details and location of…
Structural details and location of I781 hotspot in the open Ca v 1.2…
Backbone angles of wild-type and…
Backbone angles of wild-type and mutant channels. (A) Backbone angle (ψ) of position…
Structural consequences of mutations on…
Structural consequences of mutations on pore helix stability revealed by MD simulations. (A)…
Pore helices of I781T/N785A and…
Pore helices of I781T/N785A and I781T/N785G double mutants. (A) Ribbon presentation of the…
Changes in hydrophobic-hydrophobic helix interactions.…
Changes in hydrophobic-hydrophobic helix interactions. (A) Ribbon presentation of pore forming S6 segments…
Functional analysis of Ca V…
Functional analysis of Ca V 1.2 mutants in positions I781 and N785. Averaged…
Evidence for membrane localization of…
Evidence for membrane localization of mutant N785A. Transiently transfected ts-A201 cells expressing wild-type…
Computational modeling of protein mutant stability: analysis and optimization of statistical potentials and structural features reveal insights into prediction model development
Background: Understanding and predicting protein stability upon point mutations has wide-spread importance in molecular biology. Several prediction models have been developed in the past with various algorithms. Statistical potentials are one of the widely used algorithms for the prediction of changes in stability upon point mutations. Although the methods provide flexibility and the capability to develop an accurate and reliable prediction model, it can be achieved only by the right selection of the structural factors and optimization of their parameters for the statistical potentials. In this work, we have selected five atom classification systems and compared their efficiency for the development of amino acid atom potentials. Additionally, torsion angle potentials have been optimized to include the orientation of amino acids in such a way that altered backbone conformation in different secondary structural regions can be included for the prediction model. This study also elaborates the importance of classifying the mutations according to their solvent accessibility and secondary structure specificity. The prediction efficiency has been calculated individually for the mutations in different secondary structural regions and compared.
Results: Results show that, in addition to using an advanced atom description, stepwise regression and selection of atoms are necessary to avoid the redundancy in atom distribution and improve the reliability of the prediction model validation. Comparing to other atom classification models, Melo-Feytmans model shows better prediction efficiency by giving a high correlation of 0.85 between experimental and theoretical Delta Delta G with 84.06% of the mutations correctly predicted out of 1538 mutations. The theoretical Delta Delta G values for the mutations in partially buried beta-strands generated by the structural training dataset from PISCES gave a correlation of 0.84 without performing the Gaussian apodization of the torsion angle distribution. After the Gaussian apodization, the correlation increased to 0.92 and prediction accuracy increased from 80% to 88.89% respectively.
Conclusion: These findings were useful for the optimization of the Melo-Feytmans atom classification system and implementing them to develop the statistical potentials. It was also significant that the prediction efficiency of mutations in the partially buried beta-strands improves with the help of Gaussian apodization of the torsion angle distribution. All these comparisons and optimization techniques demonstrate their advantages as well as the restrictions for the development of the prediction model. These findings will be quite helpful not only for the protein stability prediction, but also for various structure solutions in future.