A machine learning approach for genome-wide prediction of morbid and druggable human genes based on systems-level data

Background: The genome-wide identification of both morbid genes, i.e., those genes whose mutations cause hereditary human diseases, and druggable genes, i.e., genes coding for proteins whose modulation by small molecules elicits phenotypic effects, requires experimental approaches that are time-consuming and laborious. Thus, a computational approach which could accurately predict such genes on a genome-wide scale would be invaluable for accelerating the pace of discovery of causal relationships between genes and diseases as well as the determination of druggability of gene products.Results: In this paper we propose a machine learning-based computational approach to predict morbid and druggable genes on a genome-wide scale. For this purpose, we constructed a decision tree-based meta-classifier and trained it on datasets containing, for each morbid and druggable gene, network topological features, tissue expression profile and subcellular localization data as learning attributes. This meta-classifier correctly recovered 65% of known morbid genes with a precision of 66% and correctly recovered 78% of known druggable genes with a precision of 75%. It was than used to assign morbidity and druggability scores to genes not known to be morbid and druggable and we showed a good match between these scores and literature data. Finally, we generated decision trees by training the J48 algorithm on the morbidity and druggability datasets to discover cellular rules for morbidity and druggability and, among the rules, we found that the number of regulating transcription factors and plasma membrane localization are the most important factors to morbidity and druggability, respectively.Conclusions: We were able to demonstrate that network topological features along with tissue expression profile and subcellular localization can reliably predict human morbid and druggable genes on a genome-wide scale. Moreover, by constructing decision trees based on these data, we could discover cellular rules governing morbidity and druggability.

Genome-wide prediction of childhood asthma and related phenotypes in a longitudinal birth cohort

Childhood wheezing and asthma vary greatly in clinical presentation and time course. The extent to which phenotypic variation reflects heterogeneity in disease pathways is unclear.

Substrate specifity of protein kinases and computational prediction of substrates

To ensure signalling fidelity, kinases must act only on a defined subset of cellular targets. Appreciating the basis for this substrate specificity is essential for understanding the role of an individual protein kinase in a particular cellular process. The specificity in the cell is determined by a combination of peptide specificity of the kinase (the molecular recognition of the sequence surrounding the phosphorylation site), substrate recruitment and phosphatase activity. Peptide specificity plays a crucial role and depends on the complementarity between the kinase and the substrate and therefore on their three-dimensional structures. Methods for experimental identification of kinase substrates and characterization of specificity are expensive and laborious, therefore, computational approaches are being developed to reduce the amount of experimental work required in substrate identification. We discuss the structural basis of substrate specificity of protein kinases and review the experimental and computational methods used to obtain specificity information. (c) 2005 Elsevier B.V. All rights reserved.

Molecular Breeding for Complex Adaptive Traits: How Integrating Crop Ecophysiology and Modelling Can Enhance Efficiency

Progress in crop improvement is limited by the ability to identify favourable combinations of genotypes (G) and management practices (M) in relevant target environments (E) given the resources available to search among the myriad of possible combinations. To underpin yield advance we require prediction of phenotype based on genotype. In plant breeding, traditional phenotypic selection methods have involved measuring phenotypic performance of large segregating populations in multi-environment trials and applying rigorous statistical procedures based on quantitative genetic theory to identify superior individuals. Recent developments in the ability to inexpensively and densely map/sequence genomes have facilitated a shift from the level of the individual (genotype) to the level of the genomic region. Molecular breeding strategies using genome wide prediction and genomic selection approaches have developed rapidly. However, their applicability to complex traits remains constrained by gene-gene and gene-environment interactions, which restrict the predictive power of associations of genomic regions with phenotypic responses. Here it is argued that crop ecophysiology and functional whole plant modelling can provide an effective link between molecular and organism scales and enhance molecular breeding by adding value to genetic prediction approaches. A physiological framework that facilitates dissection and modelling of complex traits can inform phenotyping methods for marker/gene detection and underpin prediction of likely phenotypic consequences of trait and genetic variation in target environments. This approach holds considerable promise for more effectively linking genotype to phenotype for complex adaptive traits. Specific examples focused on drought adaptation are presented to highlight the concepts.

Genomic approach to organ-specific growth control

Growth is a fundamental aspect of life cycle of all organisms. Body size varies highly in most animal groups, such as mammals. Moreover, growth of a multicellular organism is not uniform enlargement of size, but different body parts and organs grow to their characteristic sizes at different times. Currently very little is known about the molecular mechanisms governing this organ-specific growth. The genome sequencing projects have provided complete genomic DNA sequences of several species over the past decade. The amount of genomic sequence information, including sequence variants within species, is constantly increasing. Based on the universal genetic code, we can make sense of this sequence information as far as it codes proteins. However, less is known about the molecular mechanisms that control expression of genes, and about the variations in gene expression that underlie many pathological states in humans. This is caused in part by lack of information about the second genetic code that consists of the binding specificities of transcription factors and the combinatorial code by which transcription factor binding sites are assembled to form tissue-specific and/or ligand-regulated enhancer elements. This thesis presents a high-throughput assay for identification of transcription factor binding specificities, which were then used to measure the DNA binding profiles of transcription factors involved in growth control. We developed ‘enhancer element locator’, a computational tool, which can be used to predict functional enhancer elements. A genome-wide prediction of human and mouse enhancer elements generated a large database of enhancer elements. This database can be used to identify target genes of signaling pathways, and to predict activated transcription factors based on changes in gene expression. Predictions validated in transgenic mouse embryos revealed the presence of multiple tissue-specific enhancers in mouse c- and N-Myc genes, which has implications to organ specific growth control and tumor type specificity of oncogenes. Furthermore, we were able to locate a variation in a single nucleotide, which carries a susceptibility to colorectal cancer, to an enhancer element and propose a mechanism by which this SNP might be involved in generation of colorectal cancer.

The binding site distance test score: a robust method for the assessment of predicted protein binding sites

We propose a novel method for scoring the accuracy of protein binding site predictions – the Binding-site Distance Test (BDT) score. Recently, the Matthews Correlation Coefficient (MCC) has been used to evaluate binding site predictions, both by developers of new methods and by the assessors for the community wide prediction experiment – CASP8. Whilst being a rigorous scoring method, the MCC does not take into account the actual 3D location of the predicted residues from the observed binding site. Thus, an incorrectly predicted site that is nevertheless close to the observed binding site will obtain an identical score to the same number of nonbinding residues predicted at random. The MCC is somewhat affected by the subjectivity of determining observed binding residues and the ambiguity of choosing distance cutoffs. By contrast the BDT method produces continuous scores ranging between 0 and 1, relating to the distance between the predicted and observed residues. Residues predicted close to the binding site will score higher than those more distant, providing a better reflection of the true accuracy of predictions. The CASP8 function predictions were evaluated using both the MCC and BDT methods and the scores were compared. The BDT was found to strongly correlate with the MCC scores whilst also being less susceptible to the subjectivity of defining binding residues. We therefore suggest that this new simple score is a potentially more robust method for future evaluations of protein-ligand binding site predictions.

The Trypanosoma brucei MitoCarta and its regulation and splicing pattern during development

It has long been known that trypanosomes regulate mitochondrial biogenesis during the life cycle of the parasite; however, the mitochondrial protein inventory (MitoCarta) and its regulation remain unknown. We present a novel computational method for genome-wide prediction of mitochondrial proteins using a support vector machine-based classifier with approximately 90% prediction accuracy. Using this method, we predicted the mitochondrial localization of 468 proteins with high confidence and have experimentally verified the localization of a subset of these proteins. We then applied a recently developed parallel sequencing technology to determine the expression profiles and the splicing patterns of a total of 1065 predicted MitoCarta transcripts during the development of the parasite, and showed that 435 of the transcripts significantly changed their expressions while 630 remain unchanged in any of the three life stages analyzed. Furthermore, we identified 298 alternatively splicing events, a small subset of which could lead to dual localization of the corresponding proteins.

Accelerometry and Energy Expenditure in Obese Adults and Normal Weight Controls

Objective: To evaluate a new triaxial accelerometer device for prediction of energy expenditure, measured as VO2/kg, in obese adults and normal-weight controls during activities of daily life. Subjects and methods: Thirty-seven obese adults (Body Mass Index (BMI) 37±5.4) and seventeen controls (BMI 23±1.8) performed eight activities for 5 to 8 minutes while wearing a triaxial accelerometer on the right thigh. Simultaneously, VO2 and VCO2 were measured using a portable metabolic system. The relationship between accelerometer counts (AC) and VO2/kg was analysed using spline regression and linear mixed-effects models. Results: For all activities, VO2/kg was significantly lower in obese participants than in normalweight controls. A linear relationship between AC and VO2/kg existed only within accelerometer values from 0 to 300 counts/min, with an increase of 3.7 (95%-confidence interval (CI) 3.4 - 4.1) and 3.9 ml/min (95%-CI 3.4 - 4.3) per increase of 100 counts/min in obese and normal-weight adults, respectively. Linear modelling of the whole range yields wide prediction intervals for VO2/kg of ± 6.3 and ±7.3 ml/min in both groups. Conclusion: In obese and normal-weight adults, the use of AC for predicting energy expenditure, defined as VO2/kg, from a broad range of physical activities, characterized by varying intensities and types of muscle work, is limited.

Comparison of automated candidate gene prediction systems using genes implicated in type 2 diabetes by genome-wide association studies

Background
Automated candidate gene prediction systems allow geneticists to hone in on disease genes more rapidly by identifying the most probable candidate genes linked to the disease phenotypes under investigation. Here we assessed the ability of eight different candidate gene prediction systems to predict disease genes in intervals previously associated with type 2 diabetes by benchmarking their performance against genes implicated by recent genome-wide association studies.

Results
Using a search space of 9556 genes, all but one of the systems pruned the genome in favour of genes associated with moderate to highly significant SNPs. Of the 11 genes associated with highly significant SNPs identified by the genome-wide association studies, eight were flagged as likely candidates by at least one of the prediction systems. A list of candidates produced by a previous consensus approach did not match any of the genes implicated by 706 moderate to highly significant SNPs flagged by the genome-wide association studies. We prioritized genes associated with medium significance SNPs.

Conclusion
The study appraises the relative success of several candidate gene prediction systems against independent genetic data. Even when confronted with challengingly large intervals, the candidate gene prediction systems can successfully select likely disease genes. Furthermore, they can be used to filter statistically less-well-supported genetic data to select more likely candidates. We suggest consensus approaches fail because they penalize novel predictions made from independent underlying databases. To realize their full potential further work needs to be done on prioritization and annotation of genes.

Candidate disease gene prediction using Gentrepid: application to a genome-wide association study on coronary artery disease

Current single-locus-based analyses and candidate disease gene prediction methodologies used in genome-wide association studies (GWAS) do not capitalize on the wealth of the underlying genetic data, nor functional data available from molecular biology. Here, we analyzed GWAS data from the Wellcome Trust Case Control Consortium (WTCCC) on coronary artery disease (CAD). Gentrepid uses a multiple-locus-based approach, drawing on protein pathway- or domain-based data to make predictions. Known disease genes may be used as additional information (seeded method) or predictions can be based entirely on GWAS single nucleotide polymorphisms (SNPs) (ab initio method). We looked in detail at specific predictions made by Gentrepid for CAD and compared these with known genetic data and the scientific literature. Gentrepid was able to extract known disease genes from the candidate search space and predict plausible novel disease genes from both known and novel WTCCC-implicated loci. The disease gene candidates are consistent with known biological information. The results demonstrate that this computational approach is feasible and a valuable discovery tool for geneticists.

Prediction of genomic-enabled breeding values and genome-wide association study for feedlot average daily weight gain in Nelore cattle

Pós-graduação em Genética e Melhoramento Animal - FCAV

Theoretical studies of Membrane Proteins : Properties, Prediction Methods and Genome-wide analysis

Membrane proteins are a large and important class of proteins. They are responsible for several of the key functions in a living cell, e.g. transport of nutrients and ions, cell-cell signaling, and cell-cell adhesion. Despite their importance it has not been possible to study their structure and organization in much detail because of the difficulty to obtain 3D structures. In this thesis theoretical studies of membrane protein sequences and structures have been carried out by analyzing existing experimental data. The data comes from several sources including sequence databases, genome sequencing projects, and 3D structures. Prediction of the membrane spanning regions by hydrophobicity analysis is a key technique used in several of the studies. A novel method for this is also presented and compared to other methods. The primary questions addressed in the thesis are: What properties are common to all membrane proteins? What is the overall architecture of a membrane protein? What properties govern the integration into the membrane? How many membrane proteins are there and how are they distributed in different organisms? Several of the findings have now been backed up by experiments. An analysis of the large family of G-protein coupled receptors pinpoints differences in length and amino acid composition of loops between proteins with and without a signal peptide and also differences between extra- and intracellular loops. Known 3D structures of membrane proteins have been studied in terms of hydrophobicity, distribution of secondary structure and amino acid types, position specific residue variability, and differences between loops and membrane spanning regions. An analysis of several fully and partially sequenced genomes from eukaryotes, prokaryotes, and archaea has been carried out. Several differences in the membrane protein content between organisms were found, the most important being the total number of membrane proteins and the distribution of membrane proteins with a given number of transmembrane segments. Of the properties that were found to be similar in all organisms, the most obvious is the bias in the distribution of positive charges between the extra- and intracellular loops. Finally, an analysis of homologues to membrane proteins with known topology uncovered two related, multi-spanning proteins with opposite predicted orientations. The predicted topologies were verified experimentally, providing a first example of "divergent topology evolution".

Use of Bayesian geostatistical prediction to estimate local variations in Schistosoma haematobium infection in western Africa

Objective We aimed to predict sub-national spatial variation in numbers of people infected with Schistosoma haematobium, and associated uncertainties, in Burkina Faso, Mali and Niger, prior to implementation of national control programmes. Methods We used national field survey datasets covering a contiguous area 2,750 × 850 km, from 26,790 school-aged children (5–14 years) in 418 schools. Bayesian geostatistical models were used to predict prevalence of high and low intensity infections and associated 95% credible intervals (CrI). Numbers infected were determined by multiplying predicted prevalence by numbers of school-aged children in 1 km2 pixels covering the study area. Findings Numbers of school-aged children with low-intensity infections were: 433,268 in Burkina Faso, 872,328 in Mali and 580,286 in Niger. Numbers with high-intensity infections were: 416,009 in Burkina Faso, 511,845 in Mali and 254,150 in Niger. 95% CrIs (indicative of uncertainty) were wide; e.g. the mean number of boys aged 10–14 years infected in Mali was 140,200 (95% CrI 6200, 512,100). Conclusion National aggregate estimates for numbers infected mask important local variation, e.g. most S. haematobium infections in Niger occur in the Niger River valley. Prevalence of high-intensity infections was strongly clustered in foci in western and central Mali, north-eastern and northwestern Burkina Faso and the Niger River valley in Niger. Populations in these foci are likely to carry the bulk of the urinary schistosomiasis burden and should receive priority for schistosomiasis control. Uncertainties in predicted prevalence and numbers infected should be acknowledged and taken into consideration by control programme planners.

Empirical investigation of interactive highway safety design model accident prediction algorithm : Rural intersections

One major gap in transportation system safety management is the ability to assess the safety ramifications of design changes for both new road projects and modifications to existing roads. To fulfill this need, FHWA and its many partners are developing a safety forecasting tool, the Interactive Highway Safety Design Model (IHSDM). The tool will be used by roadway design engineers, safety analysts, and planners throughout the United States. As such, the statistical models embedded in IHSDM will need to be able to forecast safety impacts under a wide range of roadway configurations and environmental conditions for a wide range of driver populations and will need to be able to capture elements of driving risk across states. One of the IHSDM algorithms developed by FHWA and its contractors is for forecasting accidents on rural road segments and rural intersections. The methodological approach is to use predictive models for specific base conditions, with traffic volume information as the sole explanatory variable for crashes, and then to apply regional or state calibration factors and accident modification factors (AMFs) to estimate the impact on accidents of geometric characteristics that differ from the base model conditions. In the majority of past approaches, AMFs are derived from parameter estimates associated with the explanatory variables. A recent study for FHWA used a multistate database to examine in detail the use of the algorithm with the base model-AMF approach and explored alternative base model forms as well as the use of full models that included nontraffic-related variables and other approaches to estimate AMFs. That research effort is reported. The results support the IHSDM methodology.