![]() |
ИСТИНА |
Войти в систему Регистрация |
ФНКЦ РР |
||
Introduction Problem of rapid identification of tumors is a challenging problem for surgery and especially important for neurosurgery. Molecular profiling using rapid lipidomic-based MS methods could be a possible solution. Identification approaches used in such methods usually bases on database search with machine-learning based algorithms. Previously it was demonstrated what such approach could provide a relatively accurate result of identification in case if histopathology is used as a reference. But usually, especially in clinical practice direct comparison of histopathological data and data of MS-profiling is unavailable because of number of technical and methodological problems. This problem could be solved by development of novel approach for data validation based on direct comparison of data obtained for similar samples by different MS methods. Methods This work bases on analysis of database of brain tumor molecular profiles obtained using direct-spray-from tissue method developed by us previously. Database contains more than 6000 MS profiles of samples obtained from more than 300 patients. The primary identification of tumors was carried out using algorithms based on machine learning, trained on immunohistochemical data. However, after of such study, it is impossible to answer the question about the source of errors and the ambiguity of identification, since such methods do not allow, for example, to indicate the percentage of tumor cells present in the sample. To solve this problem, in this work, we use the data of MALDI-imaging for cross-validation of the results of rapid lipid profiling. Preliminary Data All the samples were collected in N. Burdenko research neurosurgery center in frame of operations on tumor resection. All analyzed tumors were collected from different regions of the tumor with preservation of information about the location of the tumor by the navigation system used in neurosurgical operations. Each analyzed sample was subdivided into three parts and each part was subjected to different analysis pipelines – immunohistopathology, direct rapid molecular profiling and MALDI imaging. MALDI imaging was analyzed primary by similarity measure mapping algorithm developed by us for identification of regions of tumor with homogeneous molecular structure. Database of molecular profiles contains mass spectrometry data annotated with molecular identifications based on lipidomic analysis (LC-MS/MS) performed previously. Also, MS-profiles database contains information on features associated with particular pathology. Finally, application of all this data to MS-images five us a possibility to identify regions of MALDI-image associated with particular pathology. In this work, bases on analysis of 23 samples of glioma tumors we demonstrate possibility of such approach to identify homogeneous and inhomogeneous regions of tumor. As a partial confirmation of the capabilities of the method, a decrease in the concentration of tumor cells when moving from the center of the tumor to the periphery is demonstrated. Novel Aspect First attempt to demonstrate the possibility of multimodal data analysis based on MALDI-imaging to identify inhomogeneous regions of brain tumors