Jump to content

Draft:Deep Visual Proteomics

From Wikipedia, the free encyclopedia
This is an old revision of this page, as edited by Toomuchcoding (talk | contribs) at 06:38, 28 February 2025 (→Limitations & Future Directions). The present address (URL) is a permanent link to this revision, which may differ significantly from the current revision.


Deep Visual Proteomics

Deep Visual Proteomics Overview

Deep Visual Proteomics (DVP) is a method used to analyze protein expression at the single-cell level while preserving spatial context within tissues[1]. It integrates AI-driven image analysis, single-cell laser microdissection, and mass spectrometry to address limitations in traditional proteomics techniques. By combining imaging and mass spectrometry, DVP enables the study of spatial proteomics, linking protein abundance to specific cell types and their microenvironments. This approach is applied in research on cellular diversity, disease progression, and biomarker identification, particularly in cancer studies and precision medicine.

Background

Traditional proteomics primarily utilizes bulk mass spectrometry (MS) to measure protein abundance across large cell populations, but this approach lacks single-cell resolution, making it difficult to capture cellular heterogeneity and variations in protein expression relevant to processes such as cancer progression[2]. High-performance liquid chromatography (HPLC), commonly used in MS workflows, aids in protein separation before analysis but does not retain spatial information and requires bulk cell samples[3].

Existing spatial proteomics techniques, including immunohistochemistry (IHC) and imaging mass spectrometry (IMS), allow visualization of protein distribution within tissues[4][5]. However, IHC is limited to detecting a predefined set of proteins, while IMS lacks the sensitivity and depth of MS-based approaches.

Deep Visual Proteomics (DVP) was developed to address these limitations by integrating AI-driven image analysis, single-cell laser microdissection, and MS[1][6]. This method enables protein profiling at the single-cell level while maintaining spatial relationships within tissues. By allowing analysis of cell-type-specific protein expression in disease progression and biomarker discovery, DVP provides insights into protein variation among different cell types and disease states within their native tissue environments.

Improvements Over Traditional Proteomics

Deep Visual Proteomics (DVP) offers two key advantages over traditional proteomics: single-cell resolution and spatial preservation. Bulk mass spectrometry (MS) techniques require pooling multiple cells, which masks differences between individual cells[7]. In contrast, DVP isolates single cells while retaining their spatial context within tissues, enabling a more precise analysis of protein expression across different cell types, particularly in disease research.

Compared to immunohistochemistry (IHC) and imaging mass spectrometry (IMS), DVP allows for unbiased protein discovery, as it does not rely on predefined antibody targets and provides greater sensitivity and depth than imaging-based approaches[1]. By incorporating deep learning for image segmentation and machine learning for cell classification, DVP automates and optimizes single-cell selection before proteomic analysis, improving efficiency and reproducibility.

AI-driven automation further enhances DVP’s scalability and accuracy compared to manual cell selection methods. The BIAS software facilitates image processing, feature extraction, and cell classification, increasing throughput and analytical precision. DVP serves as a powerful tool for studying cellular heterogeneity, disease mechanisms, and biomarker identification, with particular applications in cancer research and precision medicine.

Methods

Traditional proteomic methods, while effective for analyzing protein expression, often lack the spatial resolution needed to examine cellular heterogeneity and microenvironmental interactions within tissues. Mass spectrometry-based proteomics typically relies on bulk sample analysis, averaging protein expression across large cell populations and obscuring cell-specific differences and spatial context. Imaging-based techniques provide high-resolution visualization but are generally restricted to predefined biomarkers, limiting their ability to capture the full complexity of the proteome.

Deep Visual Proteomics (DVP) addresses these challenges by integrating high-resolution imaging, artificial intelligence-driven image analysis, and ultra-sensitive mass spectrometry to profile proteomes at the single-cell level while preserving spatial information[1]. This approach enables the isolation and proteomic analysis of individual cells or subcellular compartments within their native tissue context, providing insights into cellular heterogeneity, phenotypic states, and microenvironmental influences that traditional proteomics cannot resolve. DVP is particularly useful for studying complex biological processes such as cancer progression, immune responses, and tissue development, where spatial relationships and cellular diversity are key factors.

Sample Preparation

The first step in the Deep Visual Proteomics (DVP) workflow involves preparing biological samples to preserve structural integrity and ensure compatibility with high-resolution imaging and proteomic analysis. This process includes tissue selection, fixation, staining, and preparation for laser microdissection (LMD), while maintaining the spatial context necessary for single-cell proteomic profiling.

Tissue Selection & Fixation

Biological samples, such as cell cultures or tissue sections, are selected according to the specific objectives of the research. A notable advantage of DVP is its compatibility with formalin-fixed paraffin-embedded (FFPE) tissues, which are commonly stored in clinical biobanks. This compatibility enhances the potential for retrospective studies on patient samples, supporting research on disease progression and biomarker discovery. FFPE tissues maintain structural morphology and protein stability, making them suitable for the imaging and proteomic needs of DVP. In some instances, fresh-frozen tissues or cell culture monolayers may be used, especially when preserving labile post-translational modifications is critical.

Staining & Labeling

To enable high-resolution imaging and precise cell phenotyping, tissue sections or cell cultures undergo specific staining protocols. Depending on the study's requirements, immunohistochemical (IHC) staining or immunofluorescence (IF) is employed. Fluorescent dyes, such as DAPI, are commonly used to visualize nuclei, while cytoplasmic stains or membrane markers help define cell boundaries. Antibodies targeting specific proteins (e.g., EpCAM, SOX10, CD146) are applied to label particular cell types or states, an essential step for distinguishing between normal and pathological tissues or identifying rare cell populations. In more complex tissues, multiplexed immunofluorescence allows for the simultaneous staining of multiple biomarkers. This technique uses fluorophore-conjugated antibodies to enable multi-channel imaging, supporting detailed spatial analyses of cell populations. The staining process is carefully optimized to prevent excessive background or fluorescence quenching, ensuring that downstream laser microdissection and proteomic analysis are not compromised.

Slide Preparation

After staining, samples are mounted onto specialized glass membrane slides, such as polyethylene naphthalate (PEN)-coated slides, which are designed for laser microdissection (LMD). These slides provide the necessary structural support for accurate laser-guided cell excision while maintaining spatial integrity. For FFPE samples, deparaffinization and rehydration are carried out using graded ethanol washes to restore tissue morphology before staining[1]. Heat-induced epitope retrieval is then performed to unmask protein antigens, enhancing antibody binding during immunohistochemistry (IHC) or immunofluorescence (IF) staining. To ensure that tissue sections remain securely attached throughout imaging and microdissection, adhesion-promoting agents, such as APES coating, are applied. Finally, the slides are dried under controlled conditions to minimize residual moisture, preventing laser scattering or tissue distortion during microdissection.

By the end of this preparation stage, the biological samples are fully optimized for high-resolution imaging, maintaining intact spatial organization and clearly defined cellular structures[8].This meticulous sample preparation is essential for achieving the high fidelity and resolution that distinguishes DVP from traditional bulk proteomics, allowing for precise profiling of individual cells within their native tissue microenvironment.

High Resolution Imaging

The high-resolution imaging step in the Deep Visual Proteomics (DVP) workflow is for capturing the intricate structural and spatial details of biological samples, which serve as the foundation for downstream single-cell proteomic analysis. Using microscopy techniques, such as confocal or wide-field fluorescence microscopy, entire tissue sections or cell cultures are scanned to generate comprehensive two-dimensional (2D) or three-dimensional (3D) images[9]. This whole-slide scanning approach ensures that every region of the sample is documented at high resolution, preserving spatial relationships between cells and their microenvironment. The imaging process captures fine structural details at subcellular resolution, allowing for the clear identification of individual cells, nuclei, and even specific organelles within their native tissue context. This level of detail accurately segments cells and links their phenotypic characteristics to proteomic data[10]. The resulting high-parametric images facilitates precise single-cell isolation and enables a deeper understanding of tissue organization, cellular heterogeneity, and dynamic biological processes.

AI-Driven Image Analysis

AI-driven image analysis is a key aspect of the Deep Visual Proteomics (DVP) workflow, facilitating accurate cell segmentation, phenotypic classification, and data extraction from high-resolution images. The process begins with image preprocessing, where raw microscopy images undergo normalization, denoising, and contrast enhancement to improve clarity and consistency. These steps help reduce artifacts and ensure optimal conditions for subsequent segmentation and analysis. Once preprocessed, the images are analyzed using the BIAS (Biology Image Analysis Software) platform, which utilizes deep learning algorithms, such as nucleAIzer, for precise cell segmentation[9]. BIAS identifies key cellular structures, including nuclei, cytoplasm, and subcellular compartments, by analyzing morphological features and staining patterns. To enhance its adaptability across different tissue types and staining methods, the software is trained using both real and synthetically generated microscopy images, employing image style transfer techniques to create diverse and representative training datasets[9].

Following segmentation, the platform performs phenotypic classification using machine learning (ML) models that analyze a range of morphological features, including cell area, perimeter, and solidity, as well as biomarker expression levels. This allows for the identification of distinct cell types, functional states, and rare phenotypic subpopulations that may be important in health and disease. Data augmentation techniques further improve the robustness of these ML models by generating synthetic data, expanding the training dataset and enhancing model accuracy and generalizability[10]. This AI-driven approach streamlines the complex image analysis process while ensuring high precision in identifying cellular heterogeneity, enabling the accurate mapping of spatial proteomic landscapes in biological samples. By integrating AI into the DVP workflow, researchers can reveal subtle cellular variations and spatial dynamics that might be difficult to detect using traditional image analysis techniques.

Single-Cell Isolation via Laser Microdissection (LMD)

Automated Cell Selection

The single-cell isolation step in the Deep Visual Proteomics (DVP) workflow uses laser microdissection (LMD) to accurately excise individual cells or specific subcellular regions identified through AI-driven image analysis[1]. This process begins with automated cell selection, where the phenotypic data generated by the BIAS (Biology Image Analysis Software) platform guides the identification of cells or regions of interest for subsequent proteomic analysis. The AI-driven phenotyping ensures that only the most relevant cells—whether defined by morphological features, biomarker expression, or spatial location—are targeted, providing high specificity and accuracy in sample selection[11].

Laser Microdissection (LMD)

Laser microdissection is then carried out using the BIAS platform’s integration with specialized LMD microscopes, such as the Zeiss PALM MicroBeam or Leica LMD7. This integration ensures the precise transfer of cell contour data from the imaging stage to the microdissection system, allowing for excision of cells with sub-micron accuracy[11]. The system preserves spatial resolution down to 200 nanometers, maintaining the integrity of the isolated cells and their spatial context within the tissue[12]. The laser excises the selected cells by following the exact contours identified during image analysis, enabling the collection of individual cells or specific subcellular compartments without contamination from neighboring cells.

To further optimize the dissection process, advanced path-finding algorithms are utilized to minimize laser exposure and reduce the risk of laser-induced damage to the cells. These algorithms also help prevent sample loss and contamination, ensuring the isolated material remains intact and suitable for high-sensitivity proteomic analysis. The precise, automated nature of this single-cell isolation process enhances the throughput and reliability of DVP, allowing for the analysis of rare cell populations and complex tissue structures with high spatial accuracy. By combining AI-guided cell selection with laser microdissection, DVP achieves precision in isolating cells for proteomic profiling, distinguishing it from traditional bulk proteomic methods that lack spatial resolution.

Proteomic Analysis

The proteomic analysis stage of the Deep Visual Proteomics (DVP) workflow involves the identification and quantification of proteins from isolated single cells or specific subcellular compartments. After laser microdissection, the isolated material undergoes optimized protein extraction protocols designed for ultra-low input samples. These lysis methods are tailored to maximize protein recovery while preserving protein integrity, which is essential for achieving high sensitivity and accuracy in subsequent mass spectrometry (MS) analysis. Careful attention is given to minimizing sample loss and degradation during extraction, as even small inefficiencies can greatly impact data quality when working with limited cellular material.

Mass Spectrometry (MS)

Once proteins are extracted, mass spectrometry (MS) is employed to perform deep and quantitative proteomic profiling[13]. The DVP workflow utilizes Data-Independent Acquisition (DIA) techniques, particularly the diaPASEF (parallel accumulation–serial fragmentation combined with data-independent acquisition) method[14]. This approach maximizes proteomic depth by enabling the concurrent analysis of a broad range of peptides, improving sensitivity and quantification accuracy. The diaPASEF technique is well-suited for single-cell proteomics, where the detection of low-abundance proteins is critical for revealing subtle biological differences and rare cellular phenotypes.

To further enhance proteome coverage and reduce sample complexity, the workflow incorporates Ion Mobility Spectrometry, which introduces an additional dimension of separation based on the physical properties of peptide ions[15]. This added layer of separation improves the resolution of co-eluting peptides, allowing for the identification of a greater number of proteins, including those present in low quantities that might otherwise be missed in conventional MS analyses.

Quality control is a part of the proteomic analysis process to ensure data reliability and reproducibility. Internal standards are included in each run to monitor instrument performance and normalize data across samples[16]. Additionally, biological and technical replicates are analyzed to validate the consistency of protein quantification, with high correlations consistently observed across replicates (Pearson r > 0.95). These stringent quality control measures are essential for maintaining the accuracy and reproducibility of single-cell proteomic data, which can be particularly sensitive to variability due to the small amount of starting material.

Computational Biology

Deep-Learning for Cell Recognition

The first step in the DVP pipeline involves imaging and segmenting individual cells within a tissue or culture. High-resolution microscopy generates detailed images, which are then processed using deep learning-based segmentation models. The BIAS (Biology Image Analysis Software) platform plays a key role here, automatically detecting cell boundaries with high accuracy.[1][9]

To improve segmentation performance, and avoid laborious collection of real images, DVP employs pre-trained deep neural networks trained on synthetic microscopy images, allowing the model to adapt to different tissue types and staining techniques.[17]. The study compared BIAS against other segmentation tools like unet4nuclei, Cellpose, and CellProfiler, distinguishing single-cell features[18][19][20]. This step ensures that proteomic analysis is performed on well-defined, individual cells rather than mixed populations.

Machine-Learning for Feature Extraction

Once cells are segmented, machine learning (ML) algorithms classify them into distinct groups based on morphology, protein markers, and spatial context. This classification can be supervised, using predefined markers like FOXJ1 for ciliated cells, or unsupervised, where cells are clustered into novel subpopulations based on shared characteristics[21].

In addition to classification, ML is used for feature extraction, quantifying morphological traits such as cell size, shape, and neighborhood interactions[22]. These computational models allow DVP to pinpoint rare cell states that might be overlooked in traditional proteomics, making it useful for studying heterogeneous tissues like tumors.

Bioinformatics for Biological Interpretation

After single-cell laser microdissection, the selected cells undergo mass spectrometry (MS)-based proteomic analysis, generating large datasets that require extensive bioinformatics processing. This includes aligning protein features across samples to ensure comparability, normalizing data to correct for sample-to-sample variability, and applying data-independent acquisition (DIA) processing, specifically diaPASEF, which enhances the detection of low-abundance proteins[23].

Once proteins are identified, bioinformatics tools map them to biological pathways to interpret their function. In the study, DVP revealed mRNA splicing dysregulation in invasive melanoma and a decrease in interferon signaling as tumors progressed, highlighting its ability to track disease-related proteomic shifts[1].

By integrating spatial context with proteomic profiling, DVP has been used for understanding cell heterogeneity, cancer progression, and biomarker discovery, offering a new way to study proteins within their native tissue environment.

Applications of Deep Visual Proteomics

Deep Visual Proteomics (DVP) is a technology in biological research, clinical pathology, and personalized medicine by providing high-resolution, single-cell proteomic data while preserving spatial context. Its ability to integrate high-content imaging with ultra-sensitive proteomics has expanded its application across various scientific fields, offering insights into cellular heterogeneity, tissue architecture, and disease mechanisms that traditional proteomic approaches cannot resolve.

Cancer Research

DVP can contribute to cancer research by enabling the profiling of tumor heterogeneity, tracking cancer progression, and identifying spatially distinct tumor microenvironments. In melanoma research, DVP has been instrumental in revealing proteomic shifts associated with immune evasion and metastatic potential. By analyzing spatially distinct regions within tumor tissues, DVP uncovered key alterations in pathways such as mRNA splicing dysregulation and immune signaling suppression, which are linked to cancer progression and metastasis. Specifically, in studies of primary melanoma tissue, DVP identified the downregulation of interferon signaling and antigen presentation in invasive melanoma regions, highlighting mechanisms of immune escape and potential therapeutic targets​. Additionally, DVP has been applied to rare cancers, such as salivary gland acinic cell carcinoma, revealing disease-specific protein signatures and microenvironmental changes that were previously undetectable using conventional methods​.

Developmental Biology

In developmental biology, DVP has facilitated the mapping of developmental trajectories at single-cell resolution, providing a deeper understanding of the molecular programs guiding tissue formation and organogenesis. By retaining spatial information, DVP allows researchers to study how specific cell types emerge, differentiate, and interact within their native microenvironments. This has been particularly valuable in examining the development of complex tissues, such as epithelial layers and neural networks, where spatial cues are critical for proper tissue organization and function. The ability to link protein expression patterns with morphological changes enables the identification of key regulatory pathways driving development and offers new perspectives on how developmental disorders may arise from disruptions in these processes​.

Clinical Pathology

DVP holds potential in clinical pathology, particularly in the re-analysis of archived formalin-fixed paraffin-embedded (FFPE) tissue samples, which are routinely collected and stored in hospital biobanks. Traditional proteomic techniques often struggle with FFPE samples due to cross-linking and degradation; however, DVP overcomes these limitations, enabling retrospective studies on a wide range of clinical specimens. This capability has implications for biomarker discovery and personalized medicine, as historical patient samples can be revisited to identify new diagnostic markers, prognostic indicators, or therapeutic targets. For example, DVP has been used to analyze FFPE samples from melanoma and salivary gland carcinoma, revealing novel insights into disease progression and treatment resistance that were not apparent from conventional histopathology alone​. Its ability to extract spatially resolved proteomic data from archived samples opens new avenues for longitudinal studies and the development of precision oncology strategies.

Neuroscience & Immunology

The complex architecture of neural and immune tissues makes them particularly challenging to study using traditional bulk proteomics, which often obscures the contributions of individual cell types. DVP addresses this challenge by retaining spatial context during proteomic analysis, allowing for the detailed study of cell-cell interactions and microenvironmental influences in these intricate systems. In neuroscience, DVP enables the mapping of protein expression at single-cell resolution within brain tissue, shedding light on cellular heterogeneity, synaptic networks, and the molecular basis of neurological disorders. Similarly, in immunology, DVP has been used to study immune cell infiltration in tumor microenvironments and to explore the spatial dynamics of immune responses within lymphoid tissues. By linking proteomic data with cellular morphology and spatial localization, DVP provides a tool for investigating the molecular underpinnings of neurodegenerative diseases, immune evasion in cancer, and inflammatory processes​.

Limitations

Deep Visual Proteomics (DVP) has several limitations. One major challenge is the sample preparation process, particularly for formalin-fixed paraffin-embedded (FFPE) tissues, where protein recovery can be inconsistent[24]. While DVP offers high spatial resolution, analyzing individual cells remains more time-consuming compared to bulk proteomics. Additionally, AI-driven image segmentation and classification may introduce biases if models are not well-trained on diverse tissue types. While DVP provides deep proteome coverage, low-abundance proteins may still be difficult to detect due to mass spectrometry limitations[25]

Future advancements integrating DVP with multi-omics approaches like single-cell transcriptomics and spatial metabolomics could provide a more complete view of cellular mechanisms. As the technology evolves, DVP could play a role in clinical diagnostics and precision medicine, driving the discovery of novel biomarkers and therapeutic targets.

  1. ^ a b c d e f g h Mund, Andreas; Coscia, Fabian; Kriston, András; Hollandi, Réka; Kovács, Ferenc; Brunner, Andreas-David; Migh, Ede; Schweizer, Lisa; Santos, Alberto; Bzorek, Michael; Naimy, Soraya; Rahbek-Gjerdrum, Lise Mette; Dyring-Andersen, Beatrice; Bulkescher, Jutta; Lukas, Claudia (August 2022). "Deep Visual Proteomics defines single-cell identity and heterogeneity". Nature Biotechnology. 40 (8): 1231–1240. doi:10.1038/s41587-022-01302-5. ISSN 1087-0156. PMC 9371970. PMID 35590073.
  2. ^ Brunner, Andreas-David; Thielert, Marvin; Vasilopoulou, Catherine; Ammar, Constantin; Coscia, Fabian; Mund, Andreas; Hoerning, Ole B; Bache, Nicolai; Apalategui, Amalia; Lubeck, Markus; Richter, Sabrina; Fischer, David S; Raether, Oliver; Park, Melvin A; Meier, Florian (March 2022). "Ultra-high sensitivity mass spectrometry quantifies single-cell proteome changes upon perturbation". Molecular Systems Biology. 18 (3): e10798. doi:10.15252/msb.202110798. ISSN 1744-4292. PMC 8884154. PMID 35226415.
  3. ^ Zhang, Chenhua; Rodriguez, Elliott; Bi, Cong; Zheng, Xiwei; Suresh, Doddavenkatana; Suh, Kyungah; Li, Zhao; Elsebaei, Fawzi; Hage, David S. (2018). "High performance affinity chromatography and related separation methods for the analysis of biological and pharmaceutical agents". The Analyst. 143 (2): 374–391. Bibcode:2018Ana...143..374Z. doi:10.1039/C7AN01469D. ISSN 0003-2654. PMC 5768458. PMID 29200216.
  4. ^ Zhu, Shaobo; Schuerch, Conrad; Hunt, Jennifer (2015-01-01). "Review and Updates of Immunohistochemistry in Selected Salivary Gland and Head and Neck Tumors". Archives of Pathology & Laboratory Medicine. 139 (1): 55–66. doi:10.5858/arpa.2014-0167-RA. ISSN 1543-2165. PMID 25549144.
  5. ^ Buchberger, Amanda Rae; DeLaney, Kellen; Johnson, Jillian; Li, Lingjun (2018-01-02). "Mass Spectrometry Imaging: A Review of Emerging Advancements and Future Insights". Analytical Chemistry. 90 (1): 240–265. doi:10.1021/acs.analchem.7b04733. ISSN 0003-2700. PMC 5959842. PMID 29155564.
  6. ^ Zhao, Tongtong; Chiang, Zachary D.; Morriss, Julia W.; LaFave, Lindsay M.; Murray, Evan M.; Del Priore, Isabella; Meli, Kevin; Lareau, Caleb A.; Nadaf, Naeem M.; Li, Jilong; Earl, Andrew S.; Macosko, Evan Z.; Jacks, Tyler; Buenrostro, Jason D.; Chen, Fei (2022-01-06). "Spatial genomics enables multi-modal study of clonal heterogeneity in tissues". Nature. 601 (7891): 85–91. Bibcode:2022Natur.601...85Z. doi:10.1038/s41586-021-04217-4. ISSN 0028-0836. PMC 9301586. PMID 34912115.
  7. ^ Slavov, Nikolai (February 2021). "Single-cell protein analysis by mass spectrometry". Current Opinion in Chemical Biology. 60: 1–9. doi:10.1016/j.cbpa.2020.04.018. PMC 7767890. PMID 32599342.
  8. ^ Brunner, Andreas-David; Thielert, Marvin; Vasilopoulou, Catherine; Ammar, Constantin; Coscia, Fabian; Mund, Andreas; Hoerning, Ole B; Bache, Nicolai; Apalategui, Amalia; Lubeck, Markus; Richter, Sabrina; Fischer, David S; Raether, Oliver; Park, Melvin A; Meier, Florian (March 2022). "Ultra-high sensitivity mass spectrometry quantifies single-cell proteome changes upon perturbation". Molecular Systems Biology. 18 (3): e10798. doi:10.15252/msb.202110798. ISSN 1744-4292. PMC 8884154. PMID 35226415.
  9. ^ a b c d Hollandi, Reka; Szkalisity, Abel; Toth, Timea; Tasnadi, Ervin; Molnar, Csaba; Mathe, Botond; Grexa, Istvan; Molnar, Jozsef; Balind, Arpad; Gorbe, Mate; Kovacs, Maria; Migh, Ede; Goodman, Allen; Balassa, Tamas; Koos, Krisztian (May 2020). "nucleAIzer: A Parameter-free Deep Learning Framework for Nucleus Segmentation Using Image Style Transfer". Cell Systems. 10 (5): 453–458.e6. doi:10.1016/j.cels.2020.04.003. PMC 8247631. PMID 34222682.
  10. ^ a b Mehrizi, Rahil; Mehrjou, Arash; Alegro, Maryana; Zhao, Yi; Carbone, Benedetta; Fishwick, Carl; Vappiani, Johanna; Bi, Jing; Sanford, Siobhan (2023), Multi-omics Prediction from High-content Cellular Imaging with Deep Learning, arXiv:2306.09391
  11. ^ a b Xu, Min; Tocheva, Elitza I; Chang, Yi-Wei; Jensen, Grant J; Alber, Frank (2015). "De Novo Structural Pattern Mining in Cellular Electron Cryotomograms". Structure. 27 (4): 679–691.e14. arXiv:1512.09347. doi:10.1016/j.str.2019.01.005. PMC 7542605. PMID 30744995.
  12. ^ Mund, Andreas; Brunner, Andreas-David; Mann, Matthias (June 2022). "Unbiased spatial proteomics with single-cell resolution in tissues". Molecular Cell. 82 (12): 2335–2349. doi:10.1016/j.molcel.2022.05.022. PMID 35714588.
  13. ^ Lee, Donggeun; Jeon, Seung-Woo; Yi, Chang-Hwan; Kim, Yang-Hee; Choi, Yeeun; Lee, Sang-Hun; Cha, Jinwoong; Shim, Seung-Bo; Suh, Junho (2024), Diamond molecular balance: Revolutionizing high-resolution mass spectrometry from MDa to TDa at room temperature, arXiv:2406.01963
  14. ^ Rosenberger, Florian A.; Thielert, Marvin; Strauss, Maximilian T.; Ammar, Constantin; Mädler, Sophia C.; Schweizer, Lisa; Metousis, Andreas; Skowronek, Patricia; Wahle, Maria (2022-12-03), Spatial single-cell mass spectrometry defines zonation of the hepatocyte proteome, doi:10.1101/2022.12.03.518957, retrieved 2025-02-22
  15. ^ Driver, Taran; Ayers, Ruth; Pipkorn, Rüdiger; Cooper, Bridgette; Bachhawat, Nikhil; Patchkovskii, Serguei; Averbukh, Vitali; Klug, David R.; Marangos, Jon P. (2019), Partial covariance two-dimensional mass spectrometry for determination of biomolecular primary structure, arXiv:1904.05946
  16. ^ Zohora, Fatema Tuz; Rahman, M Ziaur; Tran, Ngoc Hieu; Xin, Lei; Shan, Baozhen; Li, Ming (2020), PointIso: Point Cloud Based Deep Learning Model for Detecting Arbitrary-Precision Peptide Features in LC-MS Map through Attention Based Segmentation, arXiv:2009.07250
  17. ^ Isola, Phillip; Zhu, Jun-Yan; Zhou, Tinghui; Efros, Alexei A. (July 2017). "Image-to-Image Translation with Conditional Adversarial Networks". 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE. pp. 5967–5976. arXiv:1611.07004. doi:10.1109/CVPR.2017.632. ISBN 978-1-5386-0457-1.
  18. ^ Caicedo, Juan C.; Goodman, Allen; Karhohs, Kyle W.; Cimini, Beth A.; Ackerman, Jeanelle; Haghighi, Marzieh; Heng, CherKeng; Becker, Tim; Doan, Minh; McQuin, Claire; Rohban, Mohammad; Singh, Shantanu; Carpenter, Anne E. (December 2019). "Nucleus segmentation across imaging experiments: the 2018 Data Science Bowl". Nature Methods. 16 (12): 1247–1253. doi:10.1038/s41592-019-0612-7. ISSN 1548-7105. PMC 6919559. PMID 31636459.
  19. ^ Stringer, Carsen; Wang, Tim; Michaelos, Michalis; Pachitariu, Marius (January 2021). "Cellpose: a generalist algorithm for cellular segmentation". Nature Methods. 18 (1): 100–106. doi:10.1038/s41592-020-01018-x. ISSN 1548-7105. PMID 33318659.
  20. ^ Carpenter, Anne E; Jones, Thouis R; Lamprecht, Michael R; Clarke, Colin; Kang, In Han; Friman, Ola; Guertin, David A; Chang, Joo Han; Lindquist, Robert A; Moffat, Jason; Golland, Polina; Sabatini, David M (2006-10-31). "CellProfiler: image analysis software for identifying and quantifying cell phenotypes". Genome Biology. 7 (10): R100. doi:10.1186/gb-2006-7-10-r100. ISSN 1474-760X. PMC 1794559. PMID 17076895.
  21. ^ Lengyel, Ernst (September 2010). "Ovarian Cancer Development and Metastasis". The American Journal of Pathology. 177 (3): 1053–1064. doi:10.2353/ajpath.2010.100105. PMC 2928939. PMID 20651229.
  22. ^ Berg, Stuart; Kutra, Dominik; Kroeger, Thorben; Straehle, Christoph N.; Kausler, Bernhard X.; Haubold, Carsten; Schiegg, Martin; Ales, Janez; Beier, Thorsten; Rudy, Markus; Eren, Kemal; Cervantes, Jaime I.; Xu, Buote; Beuttenmueller, Fynn; Wolny, Adrian (December 2019). "ilastik: interactive machine learning for (bio)image analysis". Nature Methods. 16 (12): 1226–1232. doi:10.1038/s41592-019-0582-9. ISSN 1548-7105. PMID 31570887.
  23. ^ Meier, Florian; Brunner, Andreas-David; Frank, Max; Ha, Annie; Bludau, Isabell; Voytik, Eugenia; Kaspar-Schoenefeld, Stephanie; Lubeck, Markus; Raether, Oliver; Bache, Nicolai; Aebersold, Ruedi; Collins, Ben C.; Röst, Hannes L.; Mann, Matthias (December 2020). "diaPASEF: parallel accumulation–serial fragmentation combined with data-independent acquisition". Nature Methods. 17 (12): 1229–1236. doi:10.1038/s41592-020-00998-0. ISSN 1548-7105. PMID 33257825.
  24. ^ Scicchitano, Marshall S.; Dalmas, Deidre A.; Boyce, Rogely W.; Thomas, Heath C.; Frazier, Kendall S. (September 2009). "Protein Extraction of Formalin-fixed, Paraffin-embedded Tissue Enables Robust Proteomic Profiles by Mass Spectrometry". Journal of Histochemistry & Cytochemistry. 57 (9): 849–860. doi:10.1369/jhc.2009.953497. ISSN 0022-1554. PMC 2728129. PMID 19471015.
  25. ^ Kim, Min-Sik; Zhong, Jun; Pandey, Akhilesh (March 2016). "Common errors in mass spectrometry-based analysis of post-translational modifications". Proteomics. 16 (5): 700–714. doi:10.1002/pmic.201500355. ISSN 1615-9853. PMC 5548100. PMID 26667783.
Retrieved from "https://en.wikipedia.org/w/index.php?title=Draft:Deep_Visual_Proteomics&oldid=1278056500"