On Predicting lung cancer subtypes using 'omic' data from tumor and tumor-adjacent histologically-normal tissue

HE2Gene: image-to-RNA translation via multi-task learning for spatial transcriptomics data

MOTIVATION

Tissue context and molecular profiling are commonly used measures in understanding normal development and disease pathology. In recent years, the development of spatial molecular profiling technologies (e.g. spatial resolved transcriptomics) has enabled the exploration of quantitative links between tissue morphology and gene expression. However, these technologies remain expensive and time-consuming, with subsequent analyses necessitating high-throughput pathological annotations. On the other hand, existing computational tools are limited to predicting only a few dozen to several hundred genes, and the majority of the methods are designed for bulk RNA-seq.

RESULTS

In this context, we propose HE2Gene, the first multi-task learning-based method capable of predicting tens of thousands of spot-level gene expressions along with pathological annotations from H&E-stained images. Experimental results demonstrate that HE2Gene is comparable to state-of-the-art methods and generalizes well on an external dataset without the need for re-training. Moreover, HE2Gene preserves the annotated spatial domains and has the potential to identify biomarkers. This capability facilitates cancer diagnosis and broadens its applicability to investigate gene-disease associations.

AVAILABILITY AND IMPLEMENTATION

The source code and data information has been deposited at https://github.com/Microbiods/HE2Gene.

An explainable AI-driven biomarker discovery framework for Non-Small Cell Lung Cancer classification

Non-Small Cell Lung Cancer (NSCLC) exhibits intrinsic heterogeneity at the molecular level that aids in distinguishing between its two prominent subtypes - Lung Adenocarcinoma (LUAD) and Lung Squamous Cell Carcinoma (LUSC). This paper proposes a novel explainable AI (XAI)-based deep learning framework to discover a small set of NSCLC biomarkers. The proposed framework comprises three modules - an autoencoder to shrink the input feature space, a feed-forward neural network to classify NSCLC instances into LUAD and LUSC, and a biomarker discovery module that leverages the combined network comprising the autoencoder and the feed-forward neural network. In the biomarker discovery module, XAI methods uncovered a set of 52 relevant biomarkers for NSCLC subtype classification. To evaluate the classification performance of the discovered biomarkers, multiple machine-learning models are constructed using these biomarkers. Using 10-Fold cross-validation, Multilayer Perceptron achieved an accuracy of 95.74% (±1.27) at 95% confidence interval. Further, using Drug-Gene Interaction Database, we observe that 14 of the discovered biomarkers are druggable. In addition, 28 biomarkers aid the prediction of the survivability of the patients. Out of 52 discovered biomarkers, we find that 45 biomarkers have been reported in previous studies on distinguishing between the two NSCLC subtypes. To the best of our knowledge, the remaining seven biomarkers have not yet been reported for NSCLC subtyping and could be further explored for their contribution to targeted therapy of lung cancer.

Endometrial Cancer-Adjacent Tissues Express Higher Levels of Cancer-Promoting Genes than the Matched Tumors

Molecular alterations in tumor-adjacent tissues have recently been recognized in some types of cancer. This phenomenon has not been studied in endometrial cancer. We aimed to analyze the expression of genes associated with cancer progression and metabolism in primary endometrial cancer samples and the matched tumor-adjacent tissues and in the samples of endometria from cancer-free patients with uterine leiomyomas. Paired samples of tumor-adjacent tissues and primary tumors from 49 patients with endometrial cancer (EC), samples of endometrium from 25 patients with leiomyomas of the uterus, and 4 endometrial cancer cell lines were examined by the RT-qPCR, for MYC, NR5A2, CXCR2, HMGA2, LIN28A, OCT4A, OCT4B, OCT4B1, TWIST1, STK11, SNAI1, and miR-205-5p expression. The expression levels of MYC, NR5A2, SNAI1, TWIST1, and STK11 were significantly higher in tumor-adjacent tissues than in the matched EC samples, and this difference was not influenced by the content of cancer cells in cancer-adjacent tissues. The expression of MYC, NR5A2, and SNAI1 was also higher in EC-adjacent tissues than in samples from cancer-free patients. In addition, the expression of MYC and CXCR2 in the tumor related to non-endometrioid adenocarcinoma and reduced the risk of recurrence, respectively, and higher NR5A2 expression in tumor-adjacent tissue increased the risk of death. In conclusion, tissues proximal to EC present higher levels of some cancer-promoting genes than the matched tumors. Malignant tumor-adjacent tissues carry a diagnostic potential and emerge as new promising target of anticancer therapy.

LSCDFS-MKL: A multiple kernel based method for lung squamous cell carcinomas disease-free survival prediction with pathological and genomic data

Lung squamous cell carcinoma (SCC) is a fatal disease in both male and female, for which current treatments are inadequate. Surgical resection is regarded as the cornerstone of treatment for patients with lung SCC, but even for the same stage patients, the wide spectrum of disease-free survival (DFS) times exits. Therefore, how to improve the DFS prediction performance of lung SCC becomes one major research area. In this study, we proposed a novel method called LSCDFS-MKL, which was on the basis of multiple kernel learning to predict DFS of lung SCC. In LSCDFS-MKL, we first efficiently integrated pathological images and genomic data (copy number aberration, gene expression, protein expression) from lung SCC. The results of LSCDFS-MKL between different types of data show that the features extracted from pathological images play an important role in DFS prediction of lung SCC. Then we compared our method LSCDFS-MKL with other existing methods and performance analysis indicates that LSCDFS-MKL has a significantly better performance than other prediction methods. After that, we applied the proposed method on different stage stratums and the performance demonstrates that LSCDFS-MKL remains efficient in DFS prediction of lung SCC patients. Finally, we performed LSCDFS-MKL on an independent validation dataset and the accuracy of DFS prediction achieves 100%, which is promising.

Genome-Wide Plasma Cell-Free DNA Methylation Profiling Identifies Potential Biomarkers for Lung Cancer

As a noninvasive blood testing, the detection of cell-free DNA (cfDNA) methylation in plasma has raised an increasing interest due to diagnostic applications. Although extensively used in cfDNA methylation analysis, bisulfite sequencing is less cost-effective. In this study, we investigated the cfDNA methylation patterns in lung cancer patients by MeDIP-seq. Compared with the healthy individuals, 330 differentially methylated regions (DMRs) at gene promoters were identified in lung cancer patients with 33 hypermethylated and 297 hypomethylated regions, respectively. Moreover, these hypermethylated genes were validated with the publicly available DNA methylation data, yielding a set of ten significant differentially methylated genes in lung cancer, including B3GAT2, BCAR1, HLF, HOPX, HOXD11, MIR1203, MYL9, SLC9A3R2, SYT5, and VTRNA1-3. Our study demonstrated MeDIP-seq could be effectively used for cfDNA methylation profiling and identified a set of potential biomarker genes with clinical application for lung cancer.

Application of omics in predicting anti-TNF efficacy in rheumatoid arthritis

Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by progressive joint erosion. Tumor necrosis factor (TNF) antagonists are the most widely used biological disease-modifying anti-rheumatic drug in RA. However, there continue to be one third of RA patients who have poor or no response to TNF antagonists. Following consideration of the uncertainty of therapeutic effects and the high price of TNF antagonists, it is worthy to predict the treatment responses before anti-TNF therapy. According to the comparisons between the responders and non-responders to TNF antagonists by omic technologies, such as genomics, transcriptomics, proteomics, and metabolomics, rheumatologists are eager to find significant biomarkers to predict the effect of TNF antagonists in order to optimize the personalized treatment in RA.

Drug discovery strategies in the field of tumor energy metabolism: Limitations by metabolic flexibility and metabolic resistance to chemotherapy

The search for new drugs capable of blocking the metabolic vulnerabilities of human tumors has now entered the clinical evaluation stage, but several projects already failed in phase I or phase II. In particular, very promising in vitro studies could not be translated in vivo at preclinical stage and beyond. This was the case for most glycolysis inhibitors that demonstrated systemic toxicity. A more recent example is the inhibition of glutamine catabolism in lung adenocarcinoma that failed in vivo despite a strong addiction of several cancer cell lines to glutamine in vitro. Such contradictory findings raised several questions concerning the optimization of drug discovery strategies in the field of cancer metabolism. For instance, the cell culture models in 2D or 3D might already show strong limitations to mimic the tumor micro- and macro-environment. The microenvironment of tumors is composed of cancer cells of variegated metabolic profiles, supporting local metabolic exchanges and symbiosis, but also of immune cells and stroma that further interact with and reshape cancer cell metabolism. The macroenvironment includes the different tissues of the organism, capable of exchanging signals and fueling the tumor 'a distance'. Moreover, most metabolic targets were identified from their increased expression in tumor transcriptomic studies, or from targeted analyses looking at the metabolic impact of particular oncogenes or tumor suppressors on selected metabolic pathways. Still, very few targets were identified from in vivo analyses of tumor metabolism in patients because such studies are difficult and adequate imaging methods are only currently being developed for that purpose. For instance, perfusion of patients with [¹³C]-glucose allows deciphering the metabolomics of tumors and opens a new area in the search for effective targets. Metabolic imaging with positron emission tomography and other techniques that do not involve [¹³C] can also be used to evaluate tumor metabolism and to follow the efficiency of a treatment at a preclinical or clinical stage. Relevant descriptors of tumor metabolism are now required to better stratify patients for the development of personalized metabolic medicine. In this review, we discuss the current limitations in basic research and drug discovery in the field of cancer metabolism to foster the need for more clinically relevant target identification and validation. We discuss the design of adapted drug screening assays and compound efficacy evaluation methods for the discovery of innovative anti-cancer therapeutic approaches at the level of tumor energetics. This article is part of a Special Issue entitled Mitochondria in Cancer, edited by Giuseppe Gasparre, Rodrigue Rossignol and Pierre Sonveaux.