Dayle L. Sampson

A comparison of methods for classifying clinical samples based on proteomics data: a case study for statistical and machine learning approaches.

The discovery of protein variation is an important strategy in disease diagnosis within the biological sciences. The current benchmark for elucidating information from multiple biological variables is the so called “omics” disciplines of the biological sciences. Such variability is uncovered by implementation of multivariable data mining techniques which come under two primary categories, machine learning strategies and statistical based approaches. Typically proteomic studies can produce hundreds or thousands of variables, p, per observation, n, depending on the analytical platform or method employed to generate the data. Many classification methods are limited by an n≪p constraint, and as such, require pre-treatment to reduce the dimensionality prior to classification. Recently machine learning techniques have gained popularity in the field for their ability to successfully classify unknown samples. One limitation of such methods is the lack of a functional model allowing meaningful interpretation of results in terms of the features used for classification. This is a problem that might be solved using a statistical model-based approach where not only is the importance of the individual protein explicit, they are combined into a readily interpretable classification rule without relying on a black box approach. Here we incorporate statistical dimension reduction techniques Partial Least Squares (PLS) and Principal Components Analysis (PCA) followed by both statistical and machine learning classification methods, and compared them to a popular machine learning technique, Support Vector Machines (SVM). Both PLS and SVM demonstrate strong utility for proteomic classification problems.

A fragment of the LG3 peptide of endorepellin is present in the urine of physically active mining workers: a potential marker of physical activity.

Biomarker analysis has been implemented in sports research in an attempt to monitor the effects of exertion and fatigue in athletes. This study proposed that while such biomarkers may be useful for monitoring injury risk in workers, proteomic approaches might also be utilised to identify novel exertion or injury markers. We found that urinary urea and cortisol levels were significantly elevated in mining workers following a 12 hour overnight shift. These levels failed to return to baseline over 24 h in the more active maintenance crew compared to truck drivers (operators) suggesting a lack of recovery between shifts. Use of a SELDI-TOF MS approach to detect novel exertion or injury markers revealed a spectral feature which was associated with workers in both work categories who were engaged in higher levels of physical activity. This feature was identified as the LG3 peptide, a C-terminal fragment of the anti-angiogenic/anti-tumourigenic protein endorepellin. This finding suggests that urinary LG3 peptide may be a biomarker of physical activity. It is also possible that the activity mediated release of LG3/endorepellin into the circulation may represent a biological mechanism for the known inverse association between physical activity and cancer risk/survival.

Urinary biomarkers of physical activity: candidates and clinical utility

Chronic physical inactivity is a major risk factor for a number of important lifestyle diseases, while inappropriate exposure to high physical demands is a risk factor for musculoskeletal injury and fatigue. Proteomic and metabolomic investigations of the physical activity continuum – extreme sedentariness to extremes in physical performance – offer increasing insight into the biological impacts of physical activity. Moreover, biomarkers, revealed in such studies, may have utility in the monitoring of metabolic and musculoskeletal health or recovery following injury. As a diagnostic matrix, urine is non-invasive to collect and it contains many biomolecules, which reflect both positive and negative adaptations to physical activity exposure. This review examines the utility and landscape of biomarkers of physical activity with particular reference to those found in urine.

Evaluation of haemoglobin in blister fluid as an indicator of paediatric burn wound depth

The early and accurate assessment of burns is essential to inform patient treatment regimens; however, this first critical step in clinical practice remains a challenge for specialist burns clinicians worldwide. In this regard, protein biomarkers are a potential adjunct diagnostic tool to assist experienced clinical judgement. Free circulating haemoglobin has previously shown some promise as an indicator of burn depth in a murine animal model. Using blister fluid collected from paediatric burn patients, haemoglobin abundance was measured using semi-quantitative Western blot and immunoassays. Although a trend was observed in which haemoglobin abundance increased with burn wound severity, several patient samples deviated significantly from this trend. Further, it was found that haemoglobin concentration decreased significantly when whole cells, cell debris and fibrinous matrix was removed from the blister fluid by centrifugation; although the relationship to depth was still present. Statistical analyses showed that haemoglobin abundance in the fluid was more strongly related to the time between injury and sample collection and the time taken for spontaneous re-epithelialisation. We hypothesise that prolonged exposure to the blister fluid microenvironment may result in an increased haemoglobin abundance due to erythrocyte lysis, and delayed wound healing.

The highly abundant urinary metabolite urobilin interferes with the bicinchoninic acid assay.

Estimation of total protein concentration is an essential step in any protein- or peptide-centric analysis pipeline. This study demonstrates that urobilin, a breakdown product of heme and a major constituent of urine, interferes considerably with the bicinchoninic acid (BCA) assay. This interference is probably due to the propensity of urobilin to reduce cupric ions (Cu(2+)) to cuprous ions (Cu(1+)), thus mimicking the reduction of copper by proteins, which the assay was designed to do. In addition, it is demonstrated that the Bradford assay is more resistant to the influence of urobilin and other small molecules. As such, urobilin has a strong confounding effect on the estimate of total protein concentrations obtained by BCA assay and thus this assay should not be used for urinary protein quantification. It is recommended that the Bradford assay be used instead.

Choose wisely: Network, ontology and annotation resources for the analysis of Staphylococcus aureus omics data.

Staphylococcus aureus (S. aureus) is a prominent human and livestock pathogen investigated widely using omic technologies. Critically, due to availability, low visibility or scattered resources, robust network and statistical contextualisation of the resulting data is generally under-represented. Here, we present novel meta-analyses of freely-accessible molecular network and gene ontology annotation information resources for S. aureus omics data interpretation. Furthermore, through the application of the gene ontology annotation resources we demonstrate their value and ability (or lack-there-of) to summarise and statistically interpret the emergent properties of gene expression and protein abundance changes using publically available data. This analysis provides simple metrics for network selection and demonstrates the availability and impact that gene ontology annotation selection can have on the contextualisation of bacterial omics data.

Identification of diagnostic and prognostic biomarkers to improve the management of diabetes-related ulcers

Abstract Introduction Diabetes-related ulcers are a common and severe complication of diabetes which is expected to increase in prevalence in line with projected global growth in rates of diabetes. Caring for these chronic wounds imposes a multi-billion dollar burden on the health care systems. These ulcers can prove lethal if untreated or not recognised and can lead to critical health complications. Methods To investigate underlying causes of wound chronicity, proteomic analyses of swab samples collected weekly from healing and non-healing diabetic foot ulcers was performed. Protein profiling was conducted based on Surface Enhanced Laser Desorption Ionisation Time of Flight (SELDI-TOF) mass spectrometry and statistical softwares were used to short list potential biomarkers. In addition, bottom-up proteomics was performed on healing and non-healing samples by SDS-PAGE and LC-MS/MS analysis using an AB SCIEX Triple TOF ® 5600 System. Trans-proteomic pipeline was used for data analyses and X! Tandem was used to search the database. Label-free quantitative proteomic analyses were performed using a computational tool called Abacus. Results (1) Statistical analyses of healing and non healing samples analysed on SELDI-TOF revealed 5 and 7 potential biomarkers (m/z) for samples with Texas score A1 and C1 respectively. (2) Bottom-up proteomics approach from both healing and non-healing samples identified 15 unique (healing) and 16 unique (non-healing) potential biomarkers. (3) Quantitative proteomic analyses resulted in 24 and 45 up-regulated healing candidates and 67 and 43 up-regulated nonhealing candidates for Texas score A1 and C1 wounds. (4) Relative quantification of 23 proteins related to oxidative stress has been identified through Abacus and 5/23 proteins have been validated using ELISA. Conclusions We are investigating these potential biomarkers using various biochemical, bioinformatics and statistical tools. A thorough investigation and study of the patterns may generate new protein candidates that can be used as potential prognostic and diagnostic biomarkers to improve the management of diabetic ulcers.