No slide title

Using Machine Learning Algorithms to Detect Cellular Stress
of Listeria monocytogenes from cDNA Microarray Data
Xiaoji Liu1, Urmila Basu1, Petr Miller1, Nasimeh Asgarian2, Russell Greiner2, Lynn M. McMullen1
University of Alberta, Department of Agricultural, Food and Nutritional Science, Edmonton, AB, Canada1, University of Alberta,
Department of Computing Science, Edmonton, AB, Canada2
Introduction
Materials and Methods
Conclusions
Listeria monocytogenes is a serious foodborne Treat L. monocytogenes with cclA
Expression levels of genes relevant to cell
 J48 Decision Tree was the most accurate pathogen that has the ability to form filaments under morphology and death
algorithm for predicting cefuroxime stress (96.9% certain environmental stress such as the presence Table 1: Genes ≥ 2-fold up or downregulated in L. monocytogenes accuracy with leave-one-out cross validation) of antimicrobials. Filament formation is the 08-5923 when exposed to cclA. The genes from this table, phenotypical sign of antimicrobial stress of as well as other relevant genes involved in cell division and  Both the J48 Decision Tree and Bayesian RNA isolation and integrity verification
PTS system (1, 5) such as lmo2002, lmo1973, lmo0633, Network were equally effective for predicting lmo1438 and lmo1892, were included in the dataset for the whether L. monocytogenes was under stress Microarrays are useful tools for measuring gene from carnocyclin A (90.0% accuracy with 5-fold expression of L. monocytogenes, and can be used to determine if a cell population undergoes Microarray
 Bayesian Nets and J48 Decision Tree could be Machine learning (ML) algorithms can use a dataset applied to detect the presence of cellular stress derived from microarrays to learn a classifier that in prokaryotes using data from DNA microarrays can later identify if a novel cell population is involved in a proposed biological process. While these Future Work
algorithms [including Bayesian Net, J48 Decision Tree, Random Forest and Support Vector Machine  Use J48 and Bayes Networks with in fold cross (SVM)] are often used to classify eukaryote validation to analyze microarray data from the microarray experiments, this study focuses on a Gene selection
prokaryotic application using two strains of  Examine the consistency of the performance of these algorithms in all the biological replicates of the microarray experiments Objectives
Performance of ML algorithms
 Test the performance of the algorithms with Table 2: the accuracy of various algorithms in predicting if a To explore if a machine learning algorithm can population of L. monocytogenes was under stress. various datasets containing expression values of learn a classifier that can predict if a population of genes from different signalling pathways L. monocytogenes is under stress from an Acknowledgements
 to distinguish between cefuroxime treated and Classify based on workflow shown
untreated L. monocytogenes EGE-e, based on This project was supported by funding from the Alberta Livestock below [WEKA (3)]
fluorescence intensity) for each gene from 32 References
Bergholz TM, B Bowen, M Wiedmann and KJ Boor (2012) Appl. Environ. Microbiol. to distinguish between L. monocytogenes 08- 5923 treated with carnocyclin A (cclA) and Dupuy A and RM Simon (2007) J. Natl. Cancer Inst. 99:147 – 57. untreated L. monocytogenes 08-5923, based Hall M, E Frank, G Holmes, B Pfahringer, P Reutemann and IH Witten (2009) The WEKA on expression level of 15 selected genes that were ≥ 2-fold up or down-regulated in the Nielsen PK, AZ Andersen, M Mols, S van der Veen,T Abee and BH Kallipolitis (2012) presence of cclA. Features were selected using Stasiewicz MJ, M Wiedmann, TM Bergholz (2011) Appl. Environ. Microbiol. 77:5294–5306.

Source: http://papersdb.cs.ualberta.ca/~papersdb/uploaded_files/1096/additional_RiboWest%20XL%20June%204_2012.pdf