The software for interactive evaluation of mass spectra stability and reproducibility.

SUMMARY: Mass spectrometry (MS) methods are widely used for the analysis of biological and medical samples. Recently developed methods, such as DESI, REIMS and NESI allow fast analyses without sample preparation at the cost of higher variability of spectra. In biology and medicine, MS profiles are often used with machine learning (classification, regression, etc.) algorithms and statistical analysis, which are sensitive to outliers and intraclass variability. Here, we present spectra similarity matrix (SSM) Display software, a tool for fast visual outlier detection and variance estimation in mass spectrometric profiles. The tool speeds up the process of manual spectra inspection, improves accuracy and explainability of outlier detection, and decreases the requirements to the operator experience. It was shown that the batch effect could be revealed through SSM analysis and that the SSM calculation can also be used for tuning novel ion sources concerning the quality of obtained mass spectra. AVAILABILITY AND IMPLEMENTATION: Source code, example datasets, binaries and other information are available at https://github.com/EvgenyZhvansky/R_matrix. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

[The role of lipids in the classification of astrocytoma and glioblastoma using MS tumor profiling]. / Rol' lipidov pri klassifikatsii astrotsitomy i glioblastomy pri pomoshchi mass-spektrometricheskogo profilirovaniia opukholei.

Express MS identification of biological tissues has become a much more accessible research method due to the application of direct specimen ionization at atmospheric pressure. In contrast to traditional methods of analysis employing GC-MS methods for determining the molecular composition of the analyzed objects it eliminates the influence of mutual ion suppression. Despite significant progress in the field of direct MS of biological tissues, the question of mass spectrometric profile attribution to a certain type of tissue still remains open. The use of modern machine learning methods and protocols (e.g., "random forests") enables us to trace possible relationships between the components of the sample MS profile and the result of brain tumor tissue classification (astrocytoma or glioblastoma). It has been shown that the most pronounced differences in the mass spectrometric profiles of these tumors are due to their lipid composition. Detection of statistically significant differences in lipid profiles of astrocytoma and glioblastoma may be used to perform an express test during surgery and inform the neurosurgeon what type of malignant tissue he is working with. The ability to accurately determine the boundaries of the neoplastic growth significantly improves the quality of both surgical intervention and postoperative rehabilitation, as well as the duration and quality of life of patients.

Metrics for evaluating the stability and reproducibility of mass spectra.

In this work, we demonstrate a new approach for assessing the stability and reproducibility of mass spectra obtained via ambient ionization methods. This method is suitable for both comparing experiments during which only one mass spectrum is measured and for evaluating the internal homogeneity of mass spectra collected over a period of time. The approach uses Pearson's r coefficient and the cosine measure to compare the spectra. It is based on the visualization of dissimilarities between measurements, thus leading to the analysis of dissimilarity patterns. The cosine measure and correlations are compared to obtain better metrics for spectra homogeneity. The method filters out unreliable scans to prevent the analyzed sample from being wrongly characterized. The applicability of the method is demonstrated on a set of brain tumor samples. The developed method could be employed in neurosurgical applications, where mass spectrometry is used to monitor the intraoperative tumor border.

Unified representation of high- and low-resolution spectra to facilitate application of mass spectrometric techniques in clinical practice.

The majority of research in the biomedical sciences is carried out with the highest resolution accessible to the scientist, but, in the clinic, cost constraints necessitate the use of low-resolution devices. Here, we compare high- and low-resolution direct mass spectrometry profiling data and propose a simple pre-processing technique that makes high-resolution data suitable for the development of classification and regression techniques applicable to low-resolution data, while retaining high accuracy of analysis. This work demonstrates an approach to de-noising spectra to make the same representation for both high- and low-resolution spectra. This approach uses noise threshold detection based on the Tversky index, which compares spectra with different resolutions, and minimizes the percentage of resolution-specific peaks. The presented method provides an avenue for the development of analytical algorithms using high-resolution mass spectrometry data, while applying these algorithms in the clinic using low-resolution mass spectrometers.

High-resolution mass spectra processing for the identification of different pathological tissue types of brain tumors.

The purpose of the work is to demonstrate the possibilities of identifying the different types of pathological tissue identification directly through tissue mass spectrometry. Glioblastoma parts dissected during neurosurgical operation were investigated. Tumor fragments were investigated by the immunohistochemistry method and were identified as necrotic tissue with necrotized vessels, necrotic tissue with tumor stain, tumor with necrosis (tumor tissue as major), tumor, necrotized tumor (necrotic tissues as major), parts of tumor cells, boundary brain tissue, and brain tissue hyperplasia. The technique of classification of tumor tissues based on mass spectrometric profile data processing is suggested in this paper. Classifiers dividing the researched sample to the corresponding tissue type were created as a result of the processing. Classifiers of necrotic and tumor tissues are shown to yield a combined result when the tissue is heterogeneous and consists of both tumor cells and necrotic tissue.