Classification of bacterial species from proteomic data using combinatorial approaches incorporating artificial neural networks, cluster analysis and principal components analysis

Lancashire, Lee; Schmid, O.; Shah, Haroun N.

doi:10.1093/bioinformatics/bti368

Cited by 62 publications

(34 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is unclear how to interpret statistical results when each protein is split across several principal components. Other approaches skip hypothesis testing by applying machine learning directly to processed proteomic data [28], or selected subsets of data [16,17], or even skip feature detection entirely by using machine learning on raw MS data [29,30]. These approaches sometimes succeed at classification, but they are likely to miss biological effects that only result in fair classifiers, and they do not generally determine the statistical significance for individual proteins.…”

Section: Discussionmentioning

confidence: 99%

Biomarker clustering to address correlations in proteomic data

Carlson¹,

Najmi²,

Cohen³

2007

Proteomics

View full text Add to dashboard Cite

Correlated variables have been shown to confound statistical analyses in microarray experiments. The same effect applies to an even greater degree in proteomics, especially with the use of MS for parallel measurements. Biological effects such as PTM, fragmentation, and multimer formation can produce strongly correlated variables. The problem is compounded in some types of MS by technical effects such as incomplete chromatographic separation, binding to multiple surfaces, or multiple ionizations. Existing methods for dimension reduction, notably principal components analysis and related techniques, are not always satisfactory because they produce data that often lack clear biological interpretation. We propose a preprocessing algorithm that clusters highly correlated features, using the Bayes information criterion to select an optimal number of clusters. Statistical analysis of clusters, instead of individual features, benefits from lower noise, and reduces the difficulties associated with strongly correlated data. This preprocessing increases the statistical power of analyses using false discovery rate on simulated data. Strong correlations are often present in real data, and we find that clustering improves biomarker discovery in clinical SELDI-TOF-MS datasets of plasma from patients with Kawasaki disease, and bone-marrow cell extracts from patients with acute myeloid or acute lymphoblastic leukemia.

show abstract

Section: Discussionmentioning

confidence: 99%

Biomarker clustering to address correlations in proteomic data

Carlson¹,

Najmi²,

Cohen³

2007

Proteomics

View full text Add to dashboard Cite

show abstract

“…An additive stepwise approach [546] was employed to identify an optimal set of markers explaining variation in the population of each of questions explored.…”

Section: Artificial Neural Networkmentioning

confidence: 99%

“…However, the application of these methods to algorithm. The models produced by ANNs have been shown to have the ability to predict well for unseen data and have the ability to cope with complexity and nonlinearity within the dataset [545,546]. Thus ANNs have the potential to identify and model patterns in this type of data to address a particular question.…”

Section: Introductionmentioning

confidence: 99%

784 Clinical applications of molecular profiling of colorectal cancer

Kheirelseid¹,

Miller²,

Chang³

et al. 2010

European Journal of Cancer Supplements

View full text Add to dashboard Cite

“…Other more computationally expensive techniques such as forward or backwards stepwise variable selection, or selection based on Artificial Neural Networks have been proposed (see eg. [Lancashire et al, 2005]), and these may capture such interactions.…”

Section: Data Miningmentioning

confidence: 99%

Mining whole-sample mass spectrometry proteomics data for biomarkers – An overview

McDonald

Skipp

Bennell

et al. 2009

Expert Systems with Applications

View full text Add to dashboard Cite

In this paper we aim to provide a concise overview of designing and conducting an MS proteomics experiment in such a way as to allow statistical analysis that may lead to the discovery of novel biomarkers. We provide a summary of the various stages that make up such an experiment, highlighting the need for experimental goals to be decided upon in advance. We discuss issues in experimental design at the sample collection stage, and good practise for standardising protocols within the proteomics laboratory. We then describe approaches to the data mining stage of the experiment, including the processing steps that transform a raw mass spectrum into a useable form. We propose a permutation-based procedure for determining the significance of reported error rates. Finally, because of its general advantages in speed and cost, we suggest that MS proteomics may be a good candidate for an early primary screening approach to disease diagnosis, identifying areas of risk and making referrals for more specific tests without necessarily making a diagnosis in its own right. Our discussion is illustrated with examples drawn from experiments on bovine blood serum conducted in the Centre for Proteomic Research (CPR) at Southampton University.

show abstract

Classification of bacterial species from proteomic data using combinatorial approaches incorporating artificial neural networks, cluster analysis and principal components analysis

Abstract: Neuroshell 2 is commercially available from Ward Systems.

Cited by 62 publications

References 15 publications

Biomarker clustering to address correlations in proteomic data

Biomarker clustering to address correlations in proteomic data

784 Clinical applications of molecular profiling of colorectal cancer

Mining whole-sample mass spectrometry proteomics data for biomarkers – An overview

Contact Info

Product

Resources

About