Introduction

Early gastric cancer (GC) detection allows for early intervention and better patient outcomes. Yet, current diagnostic methods for early GC still need to be improved. Raman microscopy offers a noninvasive, label-free approach for a wide range of ex vivo (e.g. intraoperative tumour margin detection) and in vivo (e.g. early GC diagnosis) clinical applications. However, the specificity of Raman microscopy is limited by the high signal-to-noise ratio due to light scattering and the inherent molecular complexity of human tissues.

Aims & Methods

Aim: This study aims to identify lipidomic and proteomic tissue signatures that correlate with Raman microscopy, thereby shortlisting a more specific Raman spectrum to better differentiate GC from normal gastric tissue.
Methods: Real-time Raman spectrometry-based AI was applied to 19 consecutive patients (aged 51 - 85 years old, 14 male) with high risk for gastric dysplasia or cancer. Patients underwent high-definition white light upper endoscopy. Upon localisation of the lesion/s, the endoscopist used the SPECTRA IMDx™ probe to capture the Raman spectra (1251 features) from the dysplastic lesion and its adjacent normal mucosa. Biopsies paired to the sites of Raman readings were obtained for histopathologic examination, and lipidomic and proteomic profiling. 254 lipids (from 15 subclasses) and 4168 proteins were obtained through mass-spectrometry-based techniques. Differential abundance analysis of the lipid, protein and Raman spectra was undertaken using DESeq2 [1], which uses shrinkage estimation for dispersions and fold change to improve the stability and interpretability of estimates.

Results

Using 25 tissue samples (12 control, 13 with gastric dysplasia or cancer), we identified specific lipid species and protein markers as potential biomarkers (FDR <0.05) for GC; 45 lipid species (13 cholesterol esters, nine phosphatidylethanolamines, five diacylglycerols, five lysophosphatidylethanolamines, four hex ceramides, four phosphatidylcholines, two lysophosphatidylcholines and one sphingomyelin, one ganglioside) as well as ten proteins (PLOD2, GGPS1, C8G, COX7C, DCK, ACTBL2, EMC10, H2AC21, HDHD2). Integrating the putative lipid and protein features and the Raman spectra, only a shortlist of 56 Raman features were significantly correlated with only ten lipids (3 HexCeramides, three cholesterol esters, two diacylglycerols, one phosphatidylethanolamine and one lysophosphatidylethanolamine) and two protein features (C8G, ACTBL2).

Conclusion

Our findings demonstrate the potential of integrating Raman microscopy with lipidomics and proteomics. The potential lipid and protein signatures provide insight into the biochemical alterations associated with dysplastic transformation identified by Raman microscopy. This approach brings promise for improving the accuracy of Raman-based non-invasive GC early detection strategies.

References

1. Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). https://doi.org/10.1186/s13059-014-0550-8

Disclosure

Prof Lawrence Ho Khek Yu is the Co-founder and Chief Medical Officer of Endofotonics (SPECTRA IMDx™).

Introduction

Endoscopic diagnosis of gastric dysplasia (GD) and early gastric cancer (EGC) remains challenging. Raman spectral analysis is a biomolecular specific vibrational analytical technique which measures inelastic light scattering process within samples. This technique detects disease via biomolecular "fingerprinting", which may be applied to objectively diagnose gastric lesions.

Aims & Methods

We aim to develop a machine learning model based on Raman spectra to diagnose intestinal metaplasia (IM), GD and EGC with biomarker correlations. Raman spectra were prospectively collected prior to endoscopic resection and correlated with histopathology. GD and EGC were classified as high-risk (HR) lesions; while IM and gastritis as low-risk (LR). Two methods were employed to identify significant Raman wavenumber for group differentiation: Partial Least Squares-Discriminant Analysis with Variance Importance in Projection analysis, and a theory-based statistical approach. Next, these features were tested using XGBoost model for best accuracy. A similar approach was applied to differentiate IM versus LGD, and the model was tested with leave-one-out cross-validation (LOOCV) method for the Support Vector Machine (SVM).

Results

20 participants were recruited from April 2022 to July 2023. A total of 1961 Raman spectra were collected from 33 lesions for HR versus LR differentiation. The results are as shown in Table 1. The diagnostic accuracy and F1 score for PLS-DA VIP and theory-based ranking were similar at 75.5% and 0.8318, respectively. XGBoost testing revealed that the model attained peak performance with the top 400 wavenumbers from PLS-DA VIP sorting, and the top 200 wavenumbers from the theory-based ranking. An overlap of 192 wavenumbers emerged, and the model achieved a peak F1 Score of 0.84 and 76.9% accuracy. These 192 wavenumbers were then mapped to the Raman spectrum and can be correlated with 8 Raman peaks of biochemicals related to gastric cancer including lipids, proteins, phospholipids and leucine. For LGD versus IM, we utilized 487 Raman signals from 25 lesions. was an overlap of 100 wavenumbers with accuracy of 74.1%.
The diagnostic accuracy and F1 score for PLS-DA VIP and theory-based ranking were similar at 74.4% and 0.79 respectively. LOOCV with SVM testing revealed that the model attained peak performance with the top 200 wavenumbers from PLS-DA VIP sorting, and the top 100 wavenumbers from the theory-based ranking. An overlap of 100 wavenumbers emerged, and the model achieved a peak F1 Score of 0.80 and 74.7% accuracy.

Table 1 Current AI model (XGBoost for differentiaton high risk and low risk lesions; LOOC method with SVM performance: intestinal metaplasia ) performance with selected wavenumbers for differentiation between gastric lesions.

Conclusion

A Raman spectroscopy-based AI algorithm may be useful for real time diagnosis of GIM, GD and EGC.

Introduction

Early detection of gastric cancer (GC) is crucial for improving patient outcomes, but ongoing surveillance is equally important, given the 16% risk of recurrence or progression after initial treatment. While molecular biomarkers have shown promise for diagnosis, few have been assessed in longitudinal settings.

Aims & Methods

This study aimed to develop and validate a gene expression signature capable of detecting early GC in a screening setting and to explore its utility in predicting metachronous gastric lesions (MGLs) during post-treatment surveillance following curative endoscopic submucosal dissection (ESD). A gene expression signature was developed using RNA-sequencing data from formalin-fixed paraffin-embedded (FFPE) tissue samples, including 23 controls (histologically normal mucosa) and 28 cases (low-/high-grade dysplasia and adenocarcinoma, all submitted to ESD). Gene selection was performed using machine learning within the tidymodels framework. Selected genes were validated by real-time PCR in an expanded cohort (n = 279; 123 controls, 156 cases), which was randomly split into independent discovery and validation sets (60:40 ratio). Logistic regression models were built using gene expression data, age and sex. Predictive performance for MGL development was then assessed in the 156 cases, stratified by MGL occurrence (n = 39 with MGLs, n = 157 without), with at least three years of endoscopic follow-up after curative ESD.

Results

RNA-sequencing analysis identified a seven-gene panel with potential for early GC detection, consisting of three upregulated and four downregulated genes in patients with gastric lesions. This signature was successfully validated in the stomach adenocarcinoma cohort of The Cancer Genome Atlas (TCGA-STAD) and Genotype-Tissue Expression (GTEx) datasets, upon filtering stage I cases (n = 81, P < 0.0001). In the discovery set, the model achieved excellent performance in distinguishing lesions from normal tissue (AUC = 0.96, 95% confidence interval (CI): 0.93-0.99; sensitivity = 0.90, 95% CI: 0.79-0.96; specificity = 0.94, 95% CI: 0.85-0.99). One gene retained its expression pattern in patients with MGLs. Although the model was not predictive of MGL development, a new model based on the same signature and trained exclusively on the MGL dataset achieved promising diagnostic performance (AUC = 0.74, 95% CI: 0.59-0.88; sensitivity = 0.85, 95% CI: 0.55-0.98; specificity = 0.72, 95% CI: 0.58-0.84).

Conclusion

This study demonstrates the promise of a gene expression-based model for early GC detection in a Western population. Beyond its diagnostic value, the underlying gene signature also showed potential for predicting MGLs, with good sensitivity and moderate specificity. These findings suggest that this approach could serve a dual purpose in both the detection of gastric lesions and the individualized management of these patients.

Introduction

The asymptomatic nature of precancerous lesions, namely low- and high-grade dysplasia (LGD and HGD, respectively), constitutes a major clinical challenge that precludes a timely detection of gastric cancer (GC) in curable stages. The identification of novel and accurate biomarkers holds the promise of a better risk assessment and citizens stratification for an early diagnosis. In this exploratory study, we proposed to characterize the transcriptomic landscape of gastric lesions (GL) towards the definition of a molecular profile for early screening of GC in Western populations.

Aims & Methods

RNA-Sequencing analysis was conducted on RNA extracted from 39 formalin-fixed paraffin-embedded (FFPE) samples: 20 “normal-appearing” mucosa controls and 19 gastric lesions (GL), including LGD, HGD, and adenocarcinoma. Differential expression analysis, functional enrichment analysis, least absolute shrinkage and selection operator (Lasso) regression, and the machine-learning extreme gradient boosting (XGBoost) algorithm were employed to identify potential biomarkers and assess the predictive value of the molecular signature of the top dysregulated genes.

Results

A total of 1171 differentially expressed genes (DEGs) were identified in GL. Remarkably, 867 of these genes (74%) showed exclusive dysregulation in lesions compared to stage I GC samples from The Cancer Genome Atlas (TCGA) stomach adenocarcinoma (STAD) cohort. These unique biomarkers predominantly participate in translation, as well as metabolic and biosynthetic processes. In contrast, DEGs from TCGA-STAD cohort were primarily enriched in immunity-related processes. Notably, the eight best genes model, comprising ADAMTSL1, CLDN2, ENC1, GSR, IQNC, PER1, PRR33, and a novel transcript corresponding to a long non-coding RNA (lncRNA, ENSG00000227218), achieved an area under the curve (AUC) of 0.938 (95% CI: 0.815-1.000), with a sensitivity of 87.5% and specificity of 100% to classify GL samples.

Conclusion

This preliminary study identified over 800 DEGs exclusively in GL compared to stage I GC. Moreover, it unveiled a molecular-based signature that effectively discriminates patients with GL in an intermediate-risk population. This discovery paves the way for enhancing early GC screening strategies, particularly for those who can be treated through minimally invasive, nonsurgical procedures.

Introduction

Gastric cancer has been identified as a serious global public health problem, and early detection of gastric cancer is an effective means of curbing the risk of gastric cancer. However, accurate, efficient and economical detection of early gastric cancer (EGC) remains a challenge.

Aims & Methods

We aim to fully explore the characteristics of various endoscopic images and EGC diagnostic requirements to boost the accuracy and efficiency of deep learning-based EGC detection and enable its public availability, so as to improve the survival rate and life quality of patients. A new lightweight EGC fusion diagnosis framework based on heterogeneous neural networks (FusionWN-Net) is proposed. FusionWN-Net consists of two stages: Model construction and Decision-level fusion. In the Model construction, two lightweight EGC classification models (WLI-Net and NBI-Net) are designed for the imaging characteristics of WLI and NBI to obtain accurate unimodal diagnostic results while guaranteeing low computational costs; in the Decision-level fusion, the Dempster-Shaffer algorithm is utilized to further fuse the diagnostic results of WLI-Net and NBI-Net to reach the final diagnosis. The proposed framework is trained and evaluated on a dataset of gastric endoscopic images collected and established from 1,055 patients of six hospitals, including Fuzhou University Affiliated Provincial Hospital. On this basis, we use Android Studio to develop a portable EGC diagnosis system (PEGCS) which can run smoothly on the mobile device.

Results

Extensive experiments show that the proposed framework exhibits superior performance in terms of both computational efficiency and diagnostic reliability. WLI-Net achieves 83.31% diagnostic accuracy with only 3.182M parameters and 0.471G FLOPs, and 37.04 FPS on the mobile device; NBI-Net achieves 82.85% diagnostic accuracy with 3.596M parameters and 0.609G FLOPs, and 24.39 FPS on the mobile device; the further obtained FusionWN-Net increases the accuracy to 86.52% with 6.778M parameters and 1.080G FLOPs, reaching the FPS of 14.08 on the mobile device, which is expected to be used in clinical practice.

Conclusion

We proposed WLI-Net and NBI-Net that can be used for white light endoscopic images and narrow band endoscopic images respectively, and a lightweight and efficient multimodal framework for EGC diagnosis, which are tailored for the unique characteristics and diagnostic requirements of different endoscopic images. Besides, we established the PEGCS to put the framework into practice. This work not only takes a big step forward in improving the EGC diagnostic accuracy, but also provides a general and efficient solution for obtaining high-quality EGC diagnosis results in resource-limited application scenarios, which demonstrates significant potential of clinical application.

Disclosure

The authors declare no conflicts of interest, financial or otherwise, related to the preparation or publication of this work.

Introduction

Inflammation is one of the key hallmarks of gastric cancer (GC) development, and the arachidonic acid (AA) pathway with more than 90 members has been shown to play a significant role in this process. Recent advances in omics technology have shed light on the molecular changes in cancer, providing a foundation for the identification of candidate biomarkers for use in diagnosis, prognosis, and screening. Thus, the purpose of this study was to revisit the AA pathway in GC using a transcriptomics approach, in order to identify which genes are particularly dysregulated and thus potentially representing targetable biomarkers.

Aims & Methods

RNA-Seq gene expression data (Level 3) and clinical data of the stomach adenocarcinoma TCGA cohort (STAD), containing 414 tumor samples, was obtained using cBioportal (http://cbioportal.org) and UCSC Xena (http://xena.ucsc.edu/). Shapiro–Wilk normality test was used to determine variable normality, and statistically significant differences of gene expression levels between tumor tissue and adjacent normal tissue were evaluated using the nonparametric Wilcoxon test. Hazard ratios for prognosis assessment were calculated using a univariate Cox regression model. Differential gene expression (DGE) analysis was performed using limma and edgeR R-packages. All statistical analyses were performed using R software (version 4.2.1), and p-values < 0.05 were considered statistically significant.

Results

A total of 53 genes were found to be differentially expressed in GC tumors compared with normal adjacent tissues. Overall, 29 genes were down regulated and 24 were up regulated, with the majority being involved in prostaglandin regulation and synthesis. The top down regulated genes include CYP4B1, AKR1C2, HPGD and AKR1C. Such genes encode proteins involved in AA metabolism and prostaglandin regulation and degradation, except for PTGS1, which encodes COX-1, an enzyme associated with prostaglandin synthesis. SOX9, GPX2, PON1, andFAAH2 accounted for the top up regulated genes. In the univariate analysis, four of the top DEGs correlated with worse overall survival: CYP11A1, SOX9, PON1 and FAAH2.

Conclusion

Overall, using a transcriptomic approach, the AA pathway was shown to be dysregulated in GC, with over half of the genes exhibiting altered expression, and going beyond inflammatory biomarkers, highlighting the pivotal role this pathway has in tumor development and progression and the potential for targetable actions.