Spotting malignancies from gastric endoscopic images using deep learning.

BACKGROUND: Gastric cancer is a common kind of malignancies, with yearly occurrences exceeding one million worldwide in 2017. Typically, ulcerous and cancerous tissues develop abnormal morphologies through courses of progression. Endoscopy is a routinely adopted means for examination of gastrointestinal tract for malignancy. Early and timely detection of malignancy closely correlate with good prognosis. Repeated presentation of similar frames from gastrointestinal tract endoscopy often weakens attention for practitioners to result in true patients missed out to incur higher medical cost and unnecessary morbidity. Highly needed is an automatic means for spotting visual abnormality and prompts for attention for medical staff for more thorough examination. METHODS: We conduct classification of benign ulcer and cancer for gastrointestinal endoscopic color images using deep neural network and transfer-learning approach. Using clinical data gathered from Gil Hospital, we built a dataset comprised of 200 normal, 367 cancer, and 220 ulcer cases, and applied the inception, ResNet, and VGGNet models pretrained on ImageNet. Three classes were defined-normal, benign ulcer, and cancer, and three separate binary classifiers were built-those for normal vs cancer, normal vs ulcer, and cancer vs ulcer for the corresponding classification tasks. For each task, considering inherent randomness entailed in the deep learning process, we performed data partitioning and model building experiments 100 times and averaged the performance values. RESULTS: Areas under curves of respective receiver operating characteristics were 0.95, 0.97, and 0.85 for the three classifiers. The ResNet showed the highest level of performance. The cases involving normal, i.e., normal vs ulcer and normal vs cancer resulted in accuracies above 90%. The case of ulcer vs cancer classification resulted in a lower accuracy of 77.1%, possibly due to smaller difference in appearance than those cases involving normal. CONCLUSIONS: The overall level of performance of the proposed method was very promising to encourage applications in clinical environments. Automatic classification using deep learning technique as proposed can be used to complement manual inspection efforts for practitioners to minimize dangers of missed out positives resulting from repetitive sequence of endoscopic frames and weakening attentions.

A MALDI-MS-based quantitative targeted glycomics (MALDI-QTaG) for total N-glycan analysis.

OBJECTIVES: To develop a sensitive and quantitative method for monitoring the abnormal glycosylation of clinical and biopharmaceutical products. RESULTS: MALDI-MS-based quantitative targeted glycomics (MALDI-QTaG) was proposed for sensitive and quantitative analysis of total N-glycans. The derivatization reactions (i.e., amidation of sialic acid and incorporation of a positive charge moiety into the reducing end) dramatically increased the linearity (R(2) > 0.99) and sensitivity (limit of detection is 0.5 pmol/glycoprotein) relative to underivatized glycans. In addition, the analytical strategy was chromatographic purification-free and non-laborious process accessible to the high-throughput analyses. We used MALDI-QTaG to analyze the N-glycans of α-fetoprotein (AFP) purified from normal cord blood and HCC cell line (Huh7 cells). The total percentages of core-fucosylated AFP N-glycans from Huh7 cells and normal cord blood were 98 and 18%, respectively. CONCLUSIONS: This MALDI-MS-based glycomics technology has wide applications in many clinical and bioengineering fields requiring sensitive, quantitative and fast N-glycosylation validation.

Mass spectrometry-based N-linked glycomic profiling as a means for tracking pancreatic cancer metastasis.

The aberrant glycosylation profile on the surface of cancer cells has been recognized for its potential diagnostic value towards assessing tumor progression. In this study, we initially investigate N-glycan profiles on the surface of normal (HPDE) and cancerous (Capan-1, Panc-1, and MIA PaCa-2) pancreatic cell lines, which are from different sites of pancreatic tumor. The enzymatically deglycosylated total N-glycans are permethylated via a quantitative solid-phase method and then analyzed by using MALDI-TOF MS and MALDI-QIT-TOF MS. We demonstrate that the level of high-mannose type glycans is higher among Capan-1 cells-pancreatic cancer cells that have metastasized to the liver-than that observed among Panc-1 and MIA PaCa-2 cells-pancreatic cancer cells from the pancreas duct head and tail regions, respectively. Furthermore, the relative abundance of highly-branched sialyted N-glycans is significantly up-regulated on Panc-1 and MIA PaCa-2 pancreatic cancer cells compared to that of normal HPDE pancreas cells. Taken together, these results indicate that specific N-glycosylation profile changes in pancreatic cancer cells can be used to not only distinguish between normal and cancerous cells but also provide more information on their location and metastatic potential.

Stable isotopic labeling-based quantitative targeted glycomics (i-QTaG).

Mass spectrometry (MS) analysis combined with stable isotopic labeling is a promising method for the relative quantification of aberrant glycosylation in diseases and disorders. We developed a stable isotopic labeling-based quantitative targeted glycomics (i-QTaG) technique for the comparative and quantitative analysis of total N-glycans using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). We established the analytical procedure with the chemical derivatizations (i.e., sialic acid neutralization and stable isotopic labeling) of N-glycans using a model glycoprotein (bovine fetuin). Moreover, the i-QTaG using MALDI-TOF MS was evaluated with various molar ratios (1:1, 1:2, 1:5) of (13) C6 /(12) C6 -2-aminobenzoic acid-labeled glycans from normal human serum. Finally, this method was applied to direct comparison of the total N-glycan profiles between normal human sera (n = 8) and prostate cancer patient sera (n = 17). The intensities of the N-glycan peaks from i-QTaG method showed a good linearity (R(2) > 0.99) with the amount of the bovine fetuin glycoproteins. The ratios of relative intensity between the isotopically 2-AA labeled N-glycans were close to the theoretical molar ratios (1:1, 1:2, 1:5). We also demonstrated that the up-regulation of the Lewis antigen (~82%) in sera from prostate cancer patients. In this proof-of-concept study, we demonstrated that the i-QTaG method, which enables to achieve a reliable comparative quantitation of total N-glycans via MALDI-TOF MS analysis, has the potential to diagnose and monitor alterations in glycosylation associated with disease states or biotherapeutics.

MALDI-MS-based quantitative analysis for ketone containing homoserine lactones in Pseudomonas aeruginosa.

N-Acyl homoserine lactones (AHLs), quorum sensing molecules produced by Gram-negative bacteria, are used as important secondary metabolites for antibacterial drug development and cell-to-cell communication. Although various analytical techniques have been developed for detection and quantitation of AHLs from more complex bacterial culture media, only a few methods have been applied to AHL identification in physiological samples. Here, we developed a highly sensitive and reliable MALDI-based 3-oxo AHL quantitation method by employing Girard's reagent T (GT) to produce a permanent cationic charge state [M](+) at the ketone group of AHLs. After extracting AHLs from the supernatant of bacterial cultures using ethyl acetate, the extracts were subsequently derivatized with GT without any additional purification or desalting steps. The chemical derivatization of 3-oxo AHLs dramatically enhanced sensitivity (up to 60 000 times) by lowering the limit of detection (LOD, ∼0.5 fmol)/limit of quantitation (LOQ, ∼2.5 fmol). Additionally, the GT-derivatized 3-oxo AHLs allowed more accurate quantitative analysis from the Pseudomonas aeruginosa PAO1 culture supernatants. This method may be applied for developing high-throughput and sensitive detection methods of quorum sensing signal molecules in biofilm-related clinical applications such as virulence factor characterization and antibacterial drug development.

Comparative N-linked glycan analysis of wild-type and α1,3-galactosyltransferase gene knock-out pig fibroblasts using mass spectrometry approaches.

Carbohydrate antigens expressed on pig cells are considered to be major barriers in pig-to-human xenotransplantation. Even after α1,3-galactosyltransferase gene knock-out (GalT-KO) pigs are generated, potential non-Gal antigens are still existed. However, to the best of our knowledge there is no extensive study analyzing N-glycans expressed on the GalT-KO pig tissues or cells. Here, we identified and quantified totally 47 N-glycans from wild-type (WT) and GalT-KO pig fibroblasts using mass spectrometry. First, our results confirmed the absence of galactose-alpha-1,3-galactose (α-Gal) residue in the GalT-KO pig cells. Interestingly, we showed that the level of overall fucosylated N-glycans from GalT-KO pig fibroblasts is much higher than from WT pig fibroblasts. Moreover, the relative quantity of the N-glycolylneuraminic acid (NeuGc) antigen is slightly higher in the GalT-KO pigs. Thus, this study will contribute to a better understanding of cellular glycan alterations on GalT-KO pigs for successful xenotransplantation.

The Xeno-glycomics database (XDB): a relational database of qualitative and quantitative pig glycome repertoire.

SUMMARY: In recent years, the improvement of mass spectrometry-based glycomics techniques (i.e. highly sensitive, quantitative and high-throughput analytical tools) has enabled us to obtain a large dataset of glycans. Here we present a database named Xeno-glycomics database (XDB) that contains cell- or tissue-specific pig glycomes analyzed with mass spectrometry-based techniques, including a comprehensive pig glycan information on chemical structures, mass values, types and relative quantities. It was designed as a user-friendly web-based interface that allows users to query the database according to pig tissue/cell types or glycan masses. This database will contribute in providing qualitative and quantitative information on glycomes characterized from various pig cells/organs in xenotransplantation and might eventually provide new targets in the α1,3-galactosyltransferase gene-knock out pigs era. AVAILABILITY: The database can be accessed on the web at http://bioinformatics.snu.ac.kr/xdb.