Biology subjects

Ogrinc, N.

Publications and source records attributed to Ogrinc, N..

6 recordsLinked to original sources

bioRxiv2026

CPMsi: a data-driven Computational Pipeline for fast, large-scale Mass Spectrometry Imaging analysis

Mass Spectrometry Imaging (MSI) is increasingly used in clinical research to characterize tissue heterogeneity and identify disease-associated molecular patterns while preserving spatial context. However, translating MSI to large patient cohort studies remains challenging because, among others, existing computational tools struggle to scale with the size and complexity of such datasets. To overcome this bottleneck, we developed CPMsi, a scalable, data-driven computational pipeline for cohort-scale MSI analysis. CPMsi directly processes raw MSI data, performs global spectral alignment using a novel binning strategy, identifies regions of molecular similarity, and generates consensus spectra that compactly capture spatial molecular heterogeneity. The pipeline supports mapping of regions across consecutive tissue sections, and integration of multi-modal and multi-omics MSI acquisitions from the same section. We tested CPMsi using (i) consecutive brain MSI sections comprising ~600,000 and ~500,000 pixels with ~20,000 m/z features, (ii) multimodal brain MSI lipidomic acquired from the same tissue section in positive and negative ionization modes, and (iii) a multi-omics brain MSI dataset of lipidomic, metabolomic, and glycomic measurements. CPMsi identified regions of molecular similarity 30-times faster than vendor-provided MSI analysis software. Consensus spectra substantially reduced data complexity while preserving molecular differences between tissue regions and enabled mapping regions of similar molecular composition across consecutive tissue sections. Integration of complementary MSI modalities further revealed molecular patterns that were not resolved by individual modalities alone. Overall, CPMsi enables efficient reduction, comparison, and integration of complex MSI datasets, providing a scalable framework for cohort-scale studies, biomarker discovery, and future clinical applications.

bioinformatics↗

bioRxiv2025

Image-guided alignment of consecutive multi-modal tissue slides

Multi-modal spatial data analysis often requires precise physical alignment of consecutive tissue sections, a process that can be challenging and typically relies on shared molecular markers or image recognition techniques. Here, we introduce COAST (Consecutive multi-Omics Alignment of Spatial Tissues), a method to reliably physically align consecutive tissue sections to produce a unified multi-modal molecular dataset suitable for downstream applications. COAST relies exclusively on the images associated with spatial data, eliminating the need for common molecular features or prior annotations. We demonstrate the effectiveness of COAST using spatial transcriptomics slides, where it achieves performance comparable to established uni-modal alignment tools. Applying COAST to spatial transcriptomics and metabolomics/lipidomics tissue sections from a mouse model of ischemia reperfusion injury allowed the investigation of lipid/metabolite features of transcriptionally-defined cell types. Overall, COAST offers a streamlined and integrative solution for multi-modal spatial data alignment.

bioinformatics↗

bioRxiv2024

Development of Molecular Digital Twins Based on Ambient Ionization Mass Spectrometry Imaging for Real-Time Application in Oncological Surgery

Cancer surgery is a fundamental component of oncology treatment, its quality significantly impacts patient outcomes, influencing both relapse rates and survival. However, achieving this customization is contingent upon early collection of robust molecular data during surgery, providing accurate information for diagnosis, prognosis, and delineating surgical margins. The introduction of digital twin (DT) technology has recently opened a new era of precision and effectiveness in cancer surgery. Expanding from its successful implementations in the industrial sector, DT concept has evolved into a highly promising breakthrough in healthcare. Therefore, our study goal is on creating DT by using accurate and high-throughput molecular data obtained through mass spectrometry imaging. We developed a machine-learning-based pipeline that allow to depict infiltration of cancer cells into normal tissue that offer precise delineation of tumor margins thanks to SpiderMass. This process also enables the prediction of relative presence of bacterial strains in tumoral and healthy mammary glands.

bioengineering↗

bioRxiv2020

Robot-Assisted SpiderMass for in vivo Real-Time Topography Mass Spectrometry Imaging

Mass Spectrometry Imaging (MSI) has shown to bring invaluable information for biological and clinical applications. However, conventional MSI is generally performed ex vivo from tissue sections. Here, we develop a novel MS-based method for in vivo mass spectrometry imaging. By coupling the SpiderMass technology - that provides in vivo minimally invasive analysis - to a robotic arm of high accuracy, we demonstrate that images can be acquired from any surface by moving the laser probe above the surface. By equipping the robotic arm with a sensor, we are also able to both get the topography image of the sample surface and the molecular distribution, and then and plot back the molecular data, directly to the 3D topographical image without the need for image fusion. This is shown for the first time with the 3D topographic MS-Based whole-body imaging of a mouse. Enabling fast in vivo MSI bridged to topography pave the way for surgical applications to excision margins.

biochemistry↗

bioRxiv2020

Overall patient's survival of glioblastoma associated to molecular markers: a pan-proteomic prospective study

Molecular heterogeneities are a key feature of glioblastoma (GBM) pathology impeding patients stratification and leading to high discrepancies between patients mean survivals. Here, we established a molecular classification of GBM tumors using a pan-proteomic analysis. Then, we identified, from our proteomic data, 2 clusters of biomarkers associated with good or bad patient survival from 46 IDH wild-type GBMs. Three molecular groups have been identified and associated with systemic biology analyses. Group A tumors exhibit neurogenesis characteristics and tumorigenesis. Group B shows a strong immune cell signature and express poor prognosis markers while group C tumors are characterized by an anti-viral signature and tumor growth proteins. 124 proteins were found statistically different based on patients survival times, of which 10 are issued from alternative AltORF or non-coding RNA. After statistical analysis, a panel of markers associated to higher survival (PPP1R12A, RPS14, HSPD1 and LASP1) and another panel associated to lower survival (ALCAM, ANXA11, MAOB, IP_652563 and IGHM) has been validated by immunofluorescence. Taken together, our data will guide GBM prognosis and help to improve the current GBM classification by stratifying the patients and may open new opportunities for therapeutic development. SignificanceGlioblastoma are very heterogeneous tumors with median survivals usually inferior to 20 months. We conducted a pan-proteomics analysis of glioblastoma (GBM) in order to stratify GBM based on the molecular contained. Forty-six GBM cases were classified into three groups where proteins are involved in specific pathways i.e. the first group has a neurogenesis signature and is associated with a better prognosis while the second group of patients has an immune profile with a bad prognosis. The third group is more associated to tumorigenesis. We correlated these results with the TCGA data. Finally, we have identified 28 new prognostic markers of GBM and from these 28, a panel of 4 higher and 5 lower survival markers were validated. With these 9 markers in hand, now pathologist can stratify GBM patients and can guide the therapeutic decision. HighlightsO_LIA novel stratification of glioblastoma based on mass spectrometry was established. C_LIO_LIThree groups with different molecular features and survival were identified. C_LIO_LIThis new classification could improve prognostication and may help therapeutic options. C_LIO_LI8 prognosis markers for oncologist therapeutic decision have been validated. C_LI

cancer biology↗

bioRxiv2020

Towards CNN Representations for Small Mass Spectrometry Data Classification: From Transfer Learning to Cumulative Learning

Rapid and accurate clinical diagnosis of pathological conditions remains highly challenging. A very important component of diagnosis tool development is the design of effective classification models with Mass spectrometry (MS) data. Some popular Machine Learning (ML) approaches have been investigated for this purpose but these ML models require time-consuming preprocessing steps such as baseline correction, denoising, and spectrum alignment to remove non-sample-related data artifacts. They also depend on the tedious extraction of handcrafted features, making them unsuitable for rapid analysis. Convolutional Neural Networks (CNNs) have been found to perform well under such circumstances since they can learn efficient representations from raw data without the need for costly preprocessing. However, their effectiveness drastically decreases when the number of available training samples is small, which is a common situation in medical applications. Transfer learning strategies extend an accurate representation model learnt usually on a large dataset containing many categories, to a smaller dataset with far fewer categories. In this study, we first investigate transfer learning on a 1D-CNN we have designed to classify MS data, then we develop a new representation learning method when transfer learning is not powerful enough, as in cases of low-resolution or data heterogeneity. What we propose is to train the same model through several classification tasks over various small datasets in order to accumulate generic knowledge of what MS data are, in the resulting representation. By using rat brain data as the initial training dataset, a representation learning approach can have a classification accuracy exceeding 98% for canine sarcoma cancer cells, human ovarian cancer serums, and pathogenic microorganism biotypes in 1D clinical datasets. We show for the first time the use of cumulative representation learning using datasets generated in different biological contexts, on different organisms, in different mass ranges, with different MS ionization sources, and acquired by different instruments at different resolutions. Our approach thus proposes a promising strategy for improving MS data classification accuracy when only small numbers of samples are available as a prospective cohort. The principles demonstrated in this work could even be beneficial to other domains (astronomy, archaeology...) where training samples are scarce.

bioinformatics↗