Dataset Information

A comparison of machine learning and Bayesian modelling for molecular serotyping.

ABSTRACT: Streptococcus pneumoniae is a human pathogen that is a major cause of infant mortality. Identifying the pneumococcal serotype is an important step in monitoring the impact of vaccines used to protect against disease. Genomic microarrays provide an effective method for molecular serotyping. Previously we developed an empirical Bayesian model for the classification of serotypes from a molecular serotyping array. With only few samples available, a model driven approach was the only option. In the meanwhile, several thousand samples have been made available to us, providing an opportunity to investigate serotype classification by machine learning methods, which could complement the Bayesian model.We compare the performance of the original Bayesian model with two machine learning algorithms: Gradient Boosting Machines and Random Forests. We present our results as an example of a generic strategy whereby a preliminary probabilistic model is complemented or replaced by a machine learning classifier once enough data are available. Despite the availability of thousands of serotyping arrays, a problem encountered when applying machine learning methods is the lack of training data containing mixtures of serotypes; due to the large number of possible combinations. Most of the available training data comprises samples with only a single serotype. To overcome the lack of training data we implemented an iterative analysis, creating artificial training data of serotype mixtures by combining raw data from single serotype arrays.With the enhanced training set the machine learning algorithms out perform the original Bayesian model. However, for serotypes currently lacking sufficient training data the best performing implementation was a combination of the results of the Bayesian Model and the Gradient Boosting Machine. As well as being an effective method for classifying biological data, machine learning can also be used as an efficient method for revealing subtle biological insights, which we illustrate with an example.

SUBMITTER: Newton R

PROVIDER: S-EPMC5553679 | biostudies-other | 2017 Aug

REPOSITORIES: biostudies-other

ACCESS DATA

Similar Datasets

Project description:BackgroundMicroarrays offer great potential as a platform for molecular diagnostics, testing clinical samples for the presence of numerous biomarkers in highly multiplexed assays. In this study applied to infectious diseases, data from a microarray designed for molecular serotyping of Streptococcus pneumoniae was used, identifying the presence of any one of 91 known pneumococcal serotypes from DNA extracts. This microarray incorporated oligonucleotide probes for all known capsular polysaccharide synthesis genes and required a statistical analysis of the microarray intensity data to determine which serotype, or combination of serotypes, were present within a sample based on the combination of genes detected.ResultsWe propose an empirical Bayesian model for calculating the probabilities of combinations of serotypes from the microarray data. The model takes into consideration the dependencies between serotypes, induced by genes they have in common, and by homologous genes which, although not identical, are similar to each other in sequence. For serotypes which are very similar in capsular gene composition, extra probes are included on the microarray, providing additional information which is integrated into the Bayesian model. For each serotype combination with high probability, a second model, a Bayesian random effects model is applied to determine the relative abundance of each serotype.ConclusionsTo assess the accuracy of the proposed analysis we applied our methods to experimental data from samples containing individual serotypes and samples containing combinations of serotypes with known levels of abundance. All but two of the known serotypes of S. pneumoniae that were tested as individual samples could be uniquely determined by the Bayesian model. The model also enabled the presence of combinations of serotypes within samples to be determined. Serotypes with very low abundance within a combination of serotypes can be detected (down to 2% abundance in this study). As well as detecting the presence of serotype combinations, an approximate measure of the percentage abundance of the serotypes within the combination can be obtained.

Project description:PIP4K2A is a type II lipid kinase that catalyzed the rate-limiting step of the conversion of phosphatidylinositol-5-phosphate (PI5P) into phosphatidylinositol 4,5-bisphosphate (PI4,5P2). PIP4K2A has been intricately linked to the inhibition of various types of tumors via reactive oxygen species-mediated apoptosis, making it an important therapeutic target. In the quest of finding biologically active substances with efficient PIP4K2A inhibitory activity, machine learning algorithms were used to investigate the quantitative relationship between structures and inhibitory activities of 1,7-naphthyridine analogues. Three machine learning algorithms (MLR, ANN, and SVM) were used to develop QSAR models that can effectively predict the PIP4K2A inhibitory activity of a library of 1,7-naphthyridine analogues. The cascaded feature selection method was performed by sequential application of GFA and MP5 algorithms to identify a molecular descriptor subset that can best describe the PIP4K2A inhibitory activity of 1,7-naphthyridine analogues. PIP4K2A inhibitory activities predicted by the ML models were strongly correlated with the experimental values. The QSAR Modelling indicates that the best-performing ML model was SVM with the RBF kernel function. The SVM model performed very well in predicting PIP4K2A inhibitory activity of the 1,7-naphthyridine analogues with RTR and QEX values of 0.9845 and 0.8793 respectively. To further gain more structural insight into the origin of PIP4K2A inhibitory activity of 1,7-naphthyridine analogues, molecular docking studies were performed. The results indicate that five compounds; 15, 25, 13, 09, and 28 were found to have a high binding affinity with the receptor molecules. Hydrogen bonding, pi-pi interaction, and pi-cation interactions were found to modulate the binding interaction of the inhibitors. Although the SVM gives essentially a black-box model which cannot be readily interpreted, using SVM in tandem with MLR and ANN provides a unique perspective in building robust QSAR predictive models. The superior predictive performance of the ML models and the explanatory power of MLR models were combined to provide a unique insight into the structure-activity relationship of 1,7-naphthyridine inhibitors. This is relevant in that it provides information that can be invaluable as guidelines for the design of novel PIP4K2A inhibitors.

Project description:The epidemic increase in the incidence of Human Papilloma Virus (HPV) related Oropharyngeal Squamous Cell Carcinomas (OPSCCs) in several countries worldwide represents a significant public health concern. Although gender neutral HPV vaccination programmes are expected to cause a reduction in the incidence rates of OPSCCs, these effects will not be evident in the foreseeable future. Secondary prevention strategies are currently not feasible due to an incomplete understanding of the natural history of oral HPV infections in OPSCCs. The key parameters that govern natural history models remain largely ill-defined for HPV related OPSCCs and cannot be easily inferred from experimental data. Mathematical models have been used to estimate some of these ill-defined parameters in cervical cancer, another HPV related cancer leading to successful implementation of cancer prevention strategies. We outline a "double-Bayesian" mathematical modelling approach, whereby, a Bayesian machine learning model first estimates the probability of an individual having an oral HPV infection, given OPSCC and other covariate information. The model is then inverted using Bayes' theorem to reverse the probability relationship. We use data from the Surveillance, Epidemiology, and End Results (SEER) cancer registry, SEER Head and Neck with HPV Database and the National Health and Nutrition Examination Surveys (NHANES), representing the adult population in the United States to derive our model. The model contains 8,106 OPSCC patients of which 73.0% had an oral HPV infection. When stratified by age, sex, marital status and race/ethnicity, the model estimated a higher conditional probability for developing OPSCCs given an oral HPV infection in non-Hispanic White males and females compared to other races/ethnicities. The proposed Bayesian model represents a proof-of-concept of a natural history model of HPV driven OPSCCs and outlines a strategy for estimating the conditional probability of an individual's risk of developing OPSCC following an oral HPV infection.

Dataset Information

A comparison of machine learning and Bayesian modelling for molecular serotyping.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets