Short Abstract: CEP63 and CEP152 associate with each other to form ring like structure.CEP63-CEP152 complexed ring is essential for maintaining normal centrosomal numbers in cells.Disruption in colocalisation of these two proteins is the root cause of number of diseases, particularly microcephaly. The coordinated functionality of CEP63-CEP152 ensures proper neurodevelopment, particularly human cerebral cortex growth. The co-ordinated functioning suggests that they might coevolved during their evolutionary history. The aim of this work is to get deep insight into the coevolution aspect of these two proteins in their evolutionary history. Coevolution pattern across 51 different species has been observed with the help of sequence alignment data obtained from clustalx tool. Aligned data was further analysed using TNT to find the most parsimonious tree. MEGA was used to find the most optimal model for the dataset. Summary tree was obtained by applying Bayesian algorithm. Most parsimonious trees and Bayesian analysis tree of CEP63 and CEP152 shows that CEP63 in Chimpanzee, Gorilla, Macaque and Gibbon is closely related. CEP152 in Chimpanzee, Gorilla, Macaque and Gibbon are also closely related which suggest that they have coevolved during their evolutionary history. The coevolution pattern obtained might be the result of the parallel events occurring in evolutionary time in both genes which are related to each other. The further close tracking of the coevolution path will be beneficial for the novel understanding of link enabling the formation of CEP63-CEP152 complex ring formation whose understanding will be very important for gaining insight into the diseases associated with them.

Short Abstract: Not all hepatitis C patients respond to treatment with protease inhibitors such as telaprevir (2011), boceprevir (2011), and danoprevir (Phase III). Resistance mutations in the viral protease NS3 are believed to have a determining influence on treatment's effectiveness. Next-generation technologies such as pyrosequencing offer unprecedented resolution of the viral population structure. Emergence of resistance mutations may be tracked using serial sequencing. Long-term follow-up can tell us whether resistant strains revert to wild-type (defined as absence of resistance mutations) after treatment. The study involves 15 patients infected with HCV subtype 1a, all non-responders; 2-6 timepoints spanning a period of up to 5 years; and three protease inhibitors – telaprevir, boceprevir, and danoprevir. For each sample, the protease region of HCV NS3 was sequenced using Roche/454 pyrosequencing, yielding ~1000 reads per timepoint. Viral haplotypes are inferred using hierarchical clustering. A nearest-ancestor evolutionary network is reconstructed for each patient. Intra-host evolutionary networks demonstrate broadly coherent structure in terms of propagation of resistance mutations and variant abundances. Varied branching structures provide evidence for both strong and weak selection, consistent with the assumption that evolution will be weak over short time intervals and/or in the absence of protease inhibitor, and otherwise strong. Results suggest that reversion to wild-type (i.e. the extinction of resistant strains) does occur and takes as little as 13 months. Trials of HCV protease inhibitors are potentially valuable case studies of evolution with strong selection.

Short Abstract: i-GSEA4GWAS v2.0 (improved gene set enrichment analysis for GWAS v2.0) is a web resource for analyzing biological pathways associated with traits from genome-wide association studies (GWASs). In comparison to i-GSEA4GWAS, which applies a pathway-based analysis (PBA) approach named i-GSEA (improved gene set enrichment analysis) to GWAS SNP P-values, i-GSEA4GWAS v2.0 is featured by implementing linkage disequilibrium (LD)-based SNP pruning (LSP) to obtain independent SNPs as input to i-GSEA. The current PBA tools based on GWAS SNP P-values assume independence among input SNPs. This ignores LD information on SNPs, and thus leads to biased results. So LSP is an important step to reduce or eliminate the bias for a more objective identification of trait-associated genes and pathways. Moreover, a new function of pathway comparison is implemented in i-GSEA4GWAS v2.0 to explore common genes and pathways shared by traits to further interpret outputs of i-GSEA. To our knowledge, this is the first effort that LSP and pathway comparison are implemented in a PBA tool for GWAS. i-GSEA4GWAS v2.0 is freely available at http://gsea4gwas-v2.psych.ac.cn/.

Short Abstract: This poster is based on demenstrating the résult of the evaluation of the relations between chemokine CXCL12, previously known as stromal cell-derived factor-1 (SDF1), gene variants and childhood asthma risk and disease severity in Tunisia.
Patients and Methods
Through a case–control study design, genomic DNA samples of 122 asthmatic patients and 226 age and sex-matched controls were subjected to real-time polymerase chain reaction analysis. Statistical analyses were conducted to explore the contribution of polymorphism of the CXCL12/SDF1 gene in the susceptibility to childhood asthma.
Results
Overall, the genotype frequencies of CXCL12/SDF1 gene, were significantly different between asthmatics children and controls (p < 0.001), and also different between patients with asthma of various stages (p < 0.0001).
Conclusion
A significant association between the polymorphisms of CXCL12/SDF1 and the susceptibility to childhood asthma was demonstrated.

Short Abstract: Proteins that perform nuclear activities require mechanisms of transport to get into the nucleus. The classical nuclear import pathway is directed by special signals in cargo proteins known as nuclear localization sequences (NLS) recognized by Importin-α (Impα). While Saccharomyces has a single isoform, Drosophila has three and mammals have until six copies along their genome. Several Impα isoforms have been described displaying cargo preferences, however little information is available about relationship between the paralogue sequences along Eukaryotes. Herein we present a phylogenetic analysis of Impα family by maximum likelihood and Bayesian Inference methods. Eukarya protein sequences including fungi, plants, animals and other Eukarya were collected in GenBank based on the Uniprot identifiers for the armadillo (ARM) and Importin-β binding (IBB) domains. To avoid redundancy and ensure the use of only a set of representative data, a clustering methodology was performed to cluster proteins with identity ≥98% and retrieved only one representative protein from each cluster (totaling 339 sequences in the final alignment). In the phylogenetic trees it can be observed Impα groups representing the Eukarya kingdoms. Its remarkable an early event of gene duplication following by independent events of gene loss in Metazoa and Fungi. Another important fact is the great number of independent duplication events in Metazoa and Viridiplantae. These studies will be useful as a tool to drive Isoform-selective NLS studies. Experiments to evaluate the physiological consequences of the substrate specificities can lead to the development of new ligands that can be distinguished by different isoforms.

Short Abstract: Population genetics and association studies usually rely on a set of known variable sites that are then genotyped in subsequent samples, because it is easier to genotype than to discover the variation. This is also true for structural variation detected from sequence data. However, the genotypes at known variable sites can only be inferred with uncertainty from low coverage data. Thus, the statistical approach developed by Heng Li [2011, Bioinformatics 27(21):2987-93] to infer genotype likelihoods, test hypotheses, and estimate population parameters is more appropriate. This framework allows classical population genetic analyses without accurate genotypes. Unfortunately, the implementations of these methods are intended to analyse only single nucleotide and short indel variation, and they usually assume that the two alleles in a heterozygous individual are sampled with equal probability. This is generally false for structural variation, and could lead to large biases. We present the general-purpose tool svgem, which generates maximum likelihood estimates of allele and genotype frequencies, calculates genotype posterior probabilities, and tests for Hardy-Weinberg equilibrium and for population differences, from the numbers of times the alleles are observed in each individual. It implements an expectation-maximization method and it is applicable to single nucleotide variation and structural variation of any type, observed by either split reads or paired ends, with arbitrarily high reference bias. We test svgem with simulated and real data, and we use it to genotype 13 inversions in >500 individuals using data from the 1000 Genomes Project.

Short Abstract: In computational phylogenetics, supertree methods provide a way to reconstruct larger clades of the Tree of Life. The supertree problem can be formalized in different ways to cope with contradictory information in the input. Methods are either based on encoding the input trees in a matrix or on finding minimum cuts in a graph. We developed a polynomial supertree method called FlipCut Supertrees, which computes supertrees of high quality. FlipCut uses a graph based heuristic to solve the minimum flip problem for the matrix encoding of the input trees. It builds the supertree top down by recursively splitting the FlipCut graph. FlipCut Supertrees clearly outperforms other polynomial time supertree methods but is not as good as the current best methods Matrix Representation with Parsimony (MRP) and Superfine + MRP.

To reach the quality of Superfine + MRP we propose new weightings for the FlipCut graph. One weighting type changes the basic minimum flip optimization problem to a minimum character deletion problem. The resulting FlipCut algorithm clearly outperforms the long time gold standard MRP and is on par with the best known method Superfine + MRP. In addition, it is faster and less dependent on the input data.

To further improve the quality of the supertrees, we extend the greedy FlipCut algorithm by a beam search, keeping alive a constant number of partial solutions in each phase. The guaranteed worst-case running time of the new algorithm is still polynomial in the size of the input.

Short Abstract: Often, more than one nucleotide triplet—or codon—represents an amino acid. These synonyms show a usage bias which has been studied in many model organisms among which E.coli is probably the most prominent one. There is ongoing interest in codon usage bias as it has measurable impact on the organism: some synonymous codons lead to faster gene readout in translation and, in general, “optimal” codon usage is associated with a higher fitness. Modeling can help to go beyond heuristic approaches and can link factors affecting genetic fitness to codon usage.
In detail, codon usage bias is assumed to depend on selection and mutation. Specifically we include the following potential selection pressures in our analysis
- Fitness of mutation between amino acids depending on their chemical similarity
- Time needed for tRNA insertion during translation
- Error-rate in translation
- tRNA abundance
Inspired by the use of the quasispecies model by Archetti we can, given some estimates on how the above influences codon fitness, compute the evolution of codon usage nucleotide substitution, for the distribution of codon usage and therefore gain insight on the prominence of codon usage bias.
This will allow for a better understanding of the mechanisms affecting codon usage bias as our model allows to evaluate the relevance of potential selection pressures such as translation speed or error tolerance by minimizing the difference simulated and experimental codon usage.

Short Abstract: Background: High-throughput sequencing allows the detection and
quantification of frequencies of somatic single nucleotide variants
(SNV) in heterogeneous tumor cell populations. In some cases, the
evolutionary history and population frequency of the subclonal
lineages of tumor cells present in the sample can be reconstructed simply from the SNV
frequency measurements. But algorithms to do this reconstruction are
not available and the conditions under which reconstruction is
possible have not been described.
Results: We describe the conditions under which the evolutionary history can be uniquely reconstructed from SNV frequencies from single or multiple samples from the tumor population
and we introduce a new statistical model to infer the subclonal evolutionary structure of cancer cells from these frequencies. Our model, PhyloSub, uses a Bayesian nonparametric prior over trees that groups SNVs into major subclonal lineages. PhyloSub automatically estimates the number of lineages and their ancestry. We sample from the joint posterior distribution over trees to identify evolutionary histories and cell population frequencies that have the highest probability of generating the observed SNV frequency data. When multiple phylogenies are consistent with a given set of SNV frequencies, PhyloSub is designed to explicitly represent the uncertainty in the exact phylogeny.
Conclusion: Experiments on both simulated and real data sets demonstrate the efficacy of PhyloSub. PhyloSub can successfully infer not only simple tree structures like chains but
also tree structures with branching from single and multiple tumor samples. PhyloSub-inferred phylogenies are consistent with ground truth, where available, for these datasets.

Short Abstract: High-density oligonucleotide genotyping microarrays, especially Affymetrix SNP6 chips, are widely used for high-resolution copy number analysis. In order to identify CNVs more reliable, we have proposed a Maximum a posteriori factor analysis model called cn.FARMS. The latent variable, the factor, captures the simultaneous increase or decrease of DNA amount at neighboring chromosome locations measured by the intensity of oligonucleotide probes. This increase or decrease indicates amplification or deletion of a DNA region that is a CNV. cn.FARMS considerably reduces the false discovery rate (FDR) by combining adjacent chromosome locations to an ensemble voting (agreement of multiple measurements) instead of relying on a single measurement as other methods do.

Standard factor analysis assumes a Gaussian factor distribution which, however, is a wrong assumption for CNVs. Redon et al. 2006 showed that most CNVs affect less than three individuals out of 270 HapMap samples. These rare events are hard to detect by cn.FARMS as they would be interpreted as noise. Therefore we propose a factor analysis model with a Laplacian prior, which leads to a sparse factor distribution.

We have applied the Laplacian cn.FARMS model on the HapMap dataset to detect CNVs. We could verify most of published copy number variable regions and found new ones. However many known CNVs seem to be false positives.

Short Abstract: Genomic selection (GS) combines genotypic, phenotypic and pedigree data to increase the prediction accuracy of direct genomic values (DGVs) for economically important traits. Large numbers of animals genotyped with medium to high-density panels of genetic markers are required to reach acceptable parameter estimation accuracies. Although, genotyping prices have decreased considerably in recent years, high density SNP chips are still too expensive to be used with young seedstock or commercial cattle on a routine basis. Genotype imputation can be used to address this issue and infer missing information from low-density to higher-density SNP panels. The goal of this study was to investigate the accuracy (correct call to overall call rate) of imputing 50k SNP genotypes from 3k and 6k panels using actual data derived from a Braford population. The dataset consisted of 2091 Hereford and Braford cattle (males=882 and females=1209) born in 2008 at commercial farms from southern Brazil and genotyped with the BovineSNP50 panel (Illumina Inc.). Imputation was performed over markers from autosomal chromosomes. Animals were randomly divided into 2 groups, named reference (n=1991) and validation (n=100) animals. Imputation from 3k, 6k up to 50k markers was performed using FImpute, ignoring pedigree information. Average accuracy of correctly imputed genotypes were 86.8% and 96.5% for the 3k and 6k, respectively. Imputation results were highly accurate indicating the process has a high potential to be used as a tool to lower costs for implementing GS in Braford Cattle.

Key words: imputation, beef cattle, SNP, Braford, GS

Financial support: FAPESP, EMBRAPA, CNPQ, UNEMAT

Short Abstract: Determining a published phylogenetic tree’s fitness for reuse towards some purpose depends on the availability of certain metadata, such as the type of alignment from which it was constructed, the method used to infer it, or whether branch lengths are present. This principle motivated the Minimum Information About a Phylogenetic Analysis (MIAPA) metadata reporting standard, first articulated in 2006 by Leebens-Mack et al. While many alignment and phylogenetic inference programs output some metadata suggested by the MIAPA vision, a formal definition of what is required or recommended is still lacking, and there is no commonly followed vocabulary convention that tools and users could rely on for extracting MIAPA-relevant information from a phylogenetic tree under consideration. To address this gap, we created the MIAPA ontology, an application ontology designed for annotating phylogenetic data. The ontology is developed in OWL, and reuses concepts from several existing ontologies, including the Comparative Data Analysis Ontology (CDAO), the Information Artifact Ontology (IAO), the Software Ontology (SWO and EDAM), and the W3C Provenance Ontology (PROV). The ontology aims to implement the recommendations made by an informal draft consensus checklist. Using previously published trees, we give examples for using the ontology, and we present the challenges in translating informal, sometimes ambiguous checklist recommendations into formal ontologies. We also consider the potential of formalizing a minimum metadata reporting standard as an ontology for enabling data sharing platforms to assess, rate, and thereby incentivize metadata quality and richness of their data contents for the benefit of their users.

Short Abstract: Detecting coevolving residues within or between proteins is a major computational challenge, with applications in molecular evolution, structural biology, and functional bioinformatics. It is assumed that coevolution mostly affects neighboring residues. The mutual information approach has been widely used, but has some significant drawbacks. Consequently, there have been several attempts to build models of coevolution based on biologically and statistically sound reasoning. Although more conceptually sound than mutual information, currently these approaches can only identify correlations between changes in residues or identify coevolution between pairs of binary characters.
Here we present a maximum likelihood approach for testing specific hypotheses about molecular coevolution. Our model is based on a propensity matrix that provides an explicit description on the tendency for pairs of amino acids to co-occur. The model parameterization uses this matrix to adjust the equilibrium frequency of residue pairs over evolution. This approach allows us to test the contribution of different physicochemical features on coevolution of pairs of residues. A series of likelihood ratio tests can then be used to identify pairwise interactions within a protein, with multiple testing corrected using a false discovery rate adjustment.
We applied this model to the study of coevolution of trypsin homologues, using 7 different coevolution propensity matrices. In this way, we identified sets of coevolving residues and the diverse reasons they are likely coevolving. Some sites seem to have an unrealistic tendency to coevolve with many other sites. However, filtering them out results in enrichment for coevolving pairs in close proximity.

Short Abstract: Genetic variation in gene expression makes an important contribution to phenotypic variation and susceptibility to disease. Recently, a subset of cis-acting expression quantitative loci (eQTLs) has been found to result from polymorphisms that affect RNA stability. Here we carried out a search for trans-acting variants that influence RNA stability. We first demonstrate that differences in the activity of trans-acting factors that stabilize RNA can be detected by comparing the expression levels of long-lived and short-lived genes in high-throughput gene expression experiments. Using gene expression microarray data generated from eight HapMap3 populations, we calculated the relative expression ranks of long-lived RNAs versus short-lived RNAs in each sample. Treating this as a quantitative trait, we applied genome-wide association and identified rs6137010, with which it is strongly associated in CHB and JPT. This SNP is also a cis-eQTL for SNRPB in CHB and JPT. SNRPB is a core component of the spliceosome, and affects the expression of many RNA processing factors. We propose that a cis-eQTL of SNRPB is directly responsible for inter-individual variation in relative expression of long-lived versus short-lived genes in Asians. In support of this hypothesis, knockdown of SNRPB results in a reduction in the relative expression of long-lived versus short-lived genes. Samples with higher relative expression of long lived genes had higher relative expression of coding compared to non-coding RNA and of RNA from housekeeping compared to non-housekeeping genes, due to the lower decay rates of coding RNAs, particularly those performing housekeeping functions, compared to ncRNAs.

Short Abstract: In the search for disease-associated genes, genome wide association studies (GWAS) use high-throughput genotyping method allowing analysis of a large number of single nucleotide polymorphisms (SNPs), but most of these studies have used single-locus analysis strategies. Yet, many common diseases have complex aetiologies that may involve combinations of SNPs from different genes and, possibly, different combinations within the population of affected individuals. The search for such combinations is hindered by so extreme computational and statistical difficulties, that many researchers declared that exhaustive search for 3-way interactions in modern human GWAS will never be practical. Here we demonstrate evidence that such pessimism is unfounded.

We present a model-free analysis providing a feasible approach to exhaustive search for 3-way interactions in GWAS for complex disease. For instance, the exhaustive 3-way analysis of Celiac disease GWAS containing 310,637 SNPs and 2200 samples requires only 7 days of computation on 200 Fermi based GPU cluster, which makes such analysis practical, for the first time to our knowledge. This scales down significantly for more targeted analysis, say of a specific DNA region, or of a set of SNP associated with preselected genes. In particular, the exhaustive filtering through all triplets in ~2500 SNPs including the extended HLA region required 3 minutes on our standard PC with single GTX470 NVIDIA GPU.

The top-ranked triplets in our study are significantly associated with the disease risk, have negligible marginal and bivariate effects, and show strong replications in five investigated celiac disease cohorts.

Short Abstract: We report that the expansion of 'dangerous' gene families, defined as prone to dominant deleterious mutations, can be traced to two rounds of whole-genome duplication dating back from the onset of jawed vertebrates. We argue that this striking expansion of 'dangerous' gene families implicated in genetic disorders such as cancer is a consequence of their susceptibility to deleterious mutations and the purifying selection in post-whole-genome-duplication species.

Our data mining analyses first revealed a strong correlation between the retention of duplicates from whole-genome duplication (so-called 'ohnologs') and their susceptibility to dominant deleterious mutations in human. We also investigated an alternative hypothesis frequently invoked to account for the biased retention of ohnologs, namely the 'dosage-balance' hypothesis, that posits that ohnologs are retained because their interactions with protein partners require to maintain balanced expression levels. Our results show that the gene susceptibility to deleterious mutations is more strongly correlated than dosage-balance with the retention of human ohnologs. To go beyond mere correlations, we performed mediation analyses and quantified the direct and indirect effects of many genomic properties, such as essentiality, on the retention of ohnologs. Our results demonstrate that the retention of human ohnologs is primarily caused by their susceptibility to deleterious mutations.

This supports a non-adaptive evolutionary mechanism to account for the retention of ohnologs that hinges on the purifying selection against dominant deleterious mutations in post-whole-genome-duplication species. This is because all ohnologs have been initially acquired by speciation without the need to provide evolutionary benefit to be fixed in these populations.

Short Abstract: Despite the success of GWAS, significantly associated loci generally explain only a small fraction of total heritability and provide poor predictive accuracy. Polygenic scores using less stringent P-value thresholds do capture more heritability, however, as we will show here, they are suboptimal under realistic genetic architectures. We propose a computationally efficient method for Bayesian polygenic risk score prediction that uses GWAS summary statistics as training data. This avoids both logistical and computational difficulties associated with analysis of large genotype datasets. Our method adjusts the estimated marker effects by explicitly modeling the underlying distribution of causal effect as a Gaussian mixture. Furthermore, our model accounts for linkage disequilibrium (LD) on indirect associations at non-causal markers and the effects of sampling noise. This results in a nonlinear Bayesian shrink, generalizing the standard thresholding approach. We also derive an efficient approximate Bayesian shrink that has linear running time, GoLD (Gaussian posterior mean with LD). Notably, GoLD does not require any LD information from a reference panel. Finally, we compare our approach to the standard thresholding approach using both simulated and real disease datasets, including a coronary artery disease summary statistics (Deloukas et al. 2012) with 100,000 individuals. For reasonable genetic architectures (where 1% of all SNPs are causal) and training sample sizes (20,000 individuals), GoLD doubles the prediction accuracy (as measured by squared correlation) in simulations. The proposed Bayesian shrink allows us to achieve clinically relevant prediction accuracies with smaller sample sizes than otherwise needed.

Short Abstract: Abstract: Analysis of domain occurrence and abundance in modern proteins has recently showed that the alpha/beta architecture is probably the oldest fold design. These observations have important implications for the origins of biochemistry and for exploring structure-function relationships. Here, we address the mystery of the use of chemical mechanisms by ancestral enzymes.
To test the hypothesis of the oldest ancestral folds using the most mechanisms, we have retrieved 335 enzyme reactions from the MACiE database, and mapped them over fold age (nd value; 0 being the oldest and 1 being the youngest fold). We test this concept by tracing biocatalytic mechanisms, operating in metabolic enzymes, along a phylogenetic timeline of first appearance of the homologous superfamilies of protein structures from CATH.
Results: We find that the first two folds were responsible for introducing 41% (22/53) of the known catalytic mechanisms, and observed that over half of known mechanisms were introduced before architectural diversification over the evolutionary time. The other half of the mechanisms were invented gradually over the evolutionary timeline just after organismal diversification.
Conclusion: We find that the oldest two folds in the phylogenetic analysis of protein architectures introduced many enzyme mechanisms. The most common mechanisms were invented by these folds includes fundamental building blocks of enzyme chemistry, i.e., “Proton Transfer”, “Bimolecular Nucleophilic Addition”, and “Bimolecular Nucleophilic Substitution”.