Mining mass spectra for diagnosis and biomarker discovery of cerebral accidents

Abstract

In this paper we try to identify potential biomarkers for early stroke diagnosis using surfaceenhanced laser desorption/ionization mass spectrometry coupled with analysis tools from machine learning and data mining. Data consist of 42 specimen samples, i.e., mass spectra divided in two big categories, stroke and control specimens. Among the stroke specimens two further categories exist that correspond to ischemic and hemorrhagic stroke; in this paper we limit our data analysis to discriminating between control and stroke specimens. We performed two suites of experiments. In the first one we simply applied a number of different machine learning algorithms; in the second one we have chosen the best performing algorithm as it was determined from the first phase and coupled it with a number of different feature selection methods. The reason for this was 2-fold, first to establish whether feature selection can indeed improve performance, which in our case it did not seem to confirm, but more importantly to acquire a small list of potentially interesting biomarkers. Of the different methods explored the most promising one was support vector machines which gave us high levels of sensitivity and specificity. Finally, by analyzing the models constructed by support vector machines we produced a small set of 13 features that could be used as potential biomarkers, and which exhibited good performance both in terms of sensitivity, specificity and model stability.

Stability of feature selection algorithms: a study on high-dimensional spaces

“…The goal is to construct classification models that discriminate between healthy and diseased individuals, i.e we have two class problems. We worked with three different datasets: ovarian cancer [11], (version 8-07-02), prostate cancer [12] and an extended version of the early stroke diagnosis dataset used in [14]. All features correspond to intensities of mass values and are continuous.…”

Identification of Novel Brain Biomarkers

“…Proteomic approaches have been used with only moderate success to identify potential brain biomarkers (24)(25)(26)(27). A consortium has been formed, The Human Brain Proteome Project, to facilitate the identification of brain proteins that may be involved in disease (http://www.…”

Analysis of pattern recognition and dimensionality reduction techniques for odor biometrics

“…6 shows the good performance of the system when combining ReliefF and a linear SVM, achieving recognition rates of 85.03% in the test set when 160 features are taking into account, which implies an improvement of more than 11 times with respect to the random classifier and a reduction of the dimensionality of the problem by a factor of 4.7 (752 original features/160 selected features). Actually, ReliefF has been already successfully used in mass-spectrometry domains [45]. Regarding PCA (Fig.…”

“…Therefore, ReliefF is an appealing solution to our problem since it selects the characteristic compounds of each user, rewarding those compounds that take similar values among the samples of the user but different values in the samples belonging to other individuals. Additionally, ReliefF was chosen because (i) it has already been used in mass-spectrometry domains [45]; (ii) it is a multivariate filter feature selection algorithm which does not require many training samples to obtain reliable estimations of the feature scores; (iii) it can be easily formulated for multiclass classification problems; and (iv) it can be applied before any classification algorithm.…”