Cancer Data Research Articles

Multi-omics-based Machine Learning for the Subtype Classification of Breast Cancer

AbstractCancer is a complicated disease that produces deregulatory changes in cellular activities (such as proteins). Data from these levels must be integrated into multi-omics analyses to better understand cancer and its progression. Deep learning approaches have recently helped with multi-omics analysis of cancer data. Breast cancer is a prevalent form of cancer among women, resulting from a multitude of clinical, lifestyle, social, and economic factors. The goal of this study was to predict breast cancer using several machine learning methods. We applied the architecture for mono-omics data analysis of the Cancer Genome Atlas Breast Cancer datasets in our analytical investigation. The following classifiers were used: random forest, partial least squares, Naive Bayes, decision trees, neural networks, and Lasso regularization. They were used and evaluated using the area under the curve metric. The random forest classifier and the Lasso regularization classifier achieved the highest area under the curve values of 0.99 each. These areas under the curve values were obtained using the mono-omics data employed in this investigation. The random forest and Lasso regularization classifiers achieved the maximum prediction accuracy, showing that they are appropriate for this problem. For all mono-omics classification models used in this paper, random forest and Lasso regression offer the best results for all metrics (precision, recall, and F1 score). The integration of various risk factors in breast cancer prediction modeling can aid in early diagnosis and treatment, utilizing data collection, storage, and intelligent systems for disease management. The integration of diverse risk factors in breast cancer prediction modeling holds promise for early diagnosis and treatment. Leveraging data collection, storage, and intelligent systems can further enhance disease management strategies, ultimately contributing to improved patient outcomes.

Involving patients and the public in research using linked confidential patient information without consent: a case study on the COloRECTal cancer data Repository (CORECT-R)

Objective and ApproachIn 2017, the charity Cancer Research UK funded a UK Colorectal Cancer Intelligence Hub to create the COloRECTal cancer data Repository (CORECT-R). This was intended to be a single research data system that links all datasets across the UK relevant to colorectal cancer and makes them available for research. As the four home nations (i.e. England, Scotland, Wales and Northern Ireland) have distinct data governance policies, data access and linkage has proven challenging. It has not yet been possible to create a UK-wide resource, hence limiting the potential of the data available. To address this, the Hub has recently sought to gain new governance approvals to link data from the four home nations. As part of this process, we aimed to investigate public acceptability of using confidential patient information (CPI) without consent in CORECT-R. ResultsSeventy-five participants from a diverse background attended one of four online consultation sessions. Participants raised a number of issues and concerns, including ensuring transparency in the access and use of CPI by academic and commercial partners, and being well-informed of how to opt-out. With open communication and explanation of our approach, participants gave almost unanimous support on the intended CPI use. ImplicationsInnovation and collaboration between researchers, data custodians and the public are essential to drive responsible and meaningful uses of population data. We showed an exemplar of embedding patient and public involvement in designing data linkage research. Key areas identified from the sessions will form the basis of future consultations.

PIONEER big data platform for prostate cancer: lessons for advancing future real-world evidence research.

Prostate Cancer Diagnosis and Treatment Enhancement through the Power of Big Data in Europe (PIONEER) is a European network of excellence for big data in prostate cancer. PIONEER brings together 34 private and public stakeholders from 9 countries in one multidisciplinary research consortium with the aim of positively transforming the field of prostate cancer clinical care by answering pressing questions related to prostate cancer screening, diagnosis and treatment. PIONEER has developed a unique state-of-the-art big data analytic platform by integrating existing data sources from patients withprostate cancer. PIONEER leveraged this platform to address prioritized research questions, filling knowledge gaps in the characterization, management and core outcomes of prostate cancer across the different disease stages. The network has benefited from sustained patient and stakeholder involvement and engagement, but many challenges remain when using real-world data for big data projects. To continue to advance prostate cancer care, data need to be available, suitable methodologies should be selected and mechanisms for knowledge sharing must be in place. Now acting as the prostate cancer arm of the European Association of Urology's new endeavour, UroEvidenceHub, PIONEER maintains its goal of maximizing the potential of big data to improve prostate cancer care.

Bin2cell reconstructs cells from high resolution Visium HD data.

Visium HD by 10X Genomics is the first commercially available platform capable of capturing full scale transcriptomic data paired with a reference morphology image from archived FFPE blocks at sub-cellular resolution. However, aggregation of capture regions to single cells poses challenges. Bin2cell reconstructs cells from the highest resolution data (2 μm bins) by leveraging morphology image segmentation and gene expression information. It is compatible with established Python single cell and spatial transcriptomics software, and operates efficiently in a matter of minutes without requiring a GPU. We demonstrate improvements in downstream analysis when using the reconstructed cells over default 8 μm bins on mouse brain and human colorectal cancer data. Bin2cell is available at https://github.com/Teichlab/bin2cell, along with documentation and usage examples, and can be installed from pip. Probe design functionality is available at https://github.com/Teichlab/gene2probe.

EFFECT OF GENOTYPE AND AGE ON A DEFINED MICROBIOTA IN GNOTOBIOTIC SCID PIGLETS.

Severe combined immunodeficient (SCID) individuals lack functional T and B lymphocytes, leading to a deficient adaptive immune system. SCID pigs are a unique large animal biomedical model as they possess many similarities to humans, allowing for the collection of translatable data in regenerative medicine, cancer, and other biomedical research topics. While many studies suggest early gut microbiota development is necessary for developing the intestinal barrier and immune system, these animals are often cesarian section derived, leaving them uncolonized for normal intestinal microflora. The hypothesis was that an increase in complexity of microbiota inoculum will allow for more stability in the composition of the gut microbiota of SCID piglets. This was tested across multiple litters of SCID piglets with three different defined microbiota consortium (2-strain, 6-strain, 7-strain). All piglets received their designated defined microbiota by oral gavage immediately after birth and again 24 hours later. There was no effect of SCID genotype on the composition of the gut microbiota, but there was a significant effect due to piglet age. Additionally, all three defined microbiota consortia were deemed safe to use in SCID piglets, and the 7-strain microbiota was the most stable over time. Based on these results, the 7-strain defined microbiota will be added to the SCID pig husbandry protocol, allowing for a more reproducible model.

Construction of a PANoptosis-related Prognostic Signature for Predicting Prognosis, Tumor Microenvironment, and Immune Response in Ovarian Cancer.

The PANoptosis pathway is a recently identified mechanism of cellular death that involves the interaction and synchronization among cellular pyroptosis, apoptosis, and necrosis. More and more evidence suggests that PANoptosis is involved in the development and treatment of cancer. However, a comprehensive understanding of the influence of PANoptosis genes on prognostic value, tumor microenvironment characteristics, and therapeutic outcomes in patients with ovarian cancer (OC) remains incomplete. The present work was designed to devise a PANoptosis signature for OC prognosis and explore its potential molecular function. For this study, we obtained RNA sequencing and clinical data for ovarian cancer from the Cancer Genome Atlas (TCGA) and the GSE32062 cohort. Somatic variants of PANoptosis-related genes (PRGs) in OC were analyzed using GSCA. TCGA-OC and GSE32062 were used to construct training and validation cohorts for the model. Differential expression and correlation analyses were performed following the screening of genes with prognostic ability using univariate Cox analysis. Least Absolute Shrinkage and Selection Operator (LASSO) regression was performed to construct PRG signature based on genes that were differentially expressed and correlated with prognosis. CIBERSORT and ESTIMATE were used to analyze the relationship between the PRGs signature and immune infiltration. TIDE was used to analyze the relationship between the PRG signature and immune checkpoint genes. OncoPredict was used to analyze the relationship between the PRG signature and the drug sensitivity. Quantitative real-time PCR (qRT-PCR) was used to validate the expression of PRGs in OC. The PRG signature was constructed using three prognostic genes (AIM2, APAF1, and ZBP1) in both TCGA-OC. The results showed that the PRGs signature had an AUC of 0.521, 0.546, and 0.598 in TCGA-OC and 0.620, 0.586, and 0.579 in GSE32062 to predict to predict OS at 1-, 3-, and 5-year intervals. Furthermore, a higher PRG signature risk score was significantly associated with shorter OS (HR = 1.693, 95% CI: 1.303 - 2.202, p = 8.34 × 10^-5 in TCGA-OC and HR = 1.63, 95% CI: 1.13 - 2.35, p = 0.009 in GSE32062). The risk score was identified as the independent prognostic factor for OC. Patients categorized according to their risk score exhibited notable variations in immune status, response to immunotherapy, and sensitivity to drugs. AIM2, APAF1, and ZBP1 were significantly aberrantly expressed in OC cell lines. The PRG signature has the potential to serve as a prognostic predictor for OC and to provide new insights into OC treatment.

Impact of CDKN2A gene expression on colon adenocarcinoma via biosignature analysis.

Colorectal adenocarcinoma (COAD) has a poor prognosis. Cyclin-dependent kinase inhibitor 2A (CDKN2A) significantly affects the development and progression of various human tumors. However, the significance and pathological mechanisms of CDKN2A in COAD remain to be elucidated. We assessed expression levels, clinical significance, biological function, co-expressed genes, and enrichment of related pathways of CDKN2A in COAD using various databases, including The University of Alabama at Birmingham Cancer Data Analysis Portal, Gene Expression Profiling Interactive Analysis, Tumor Immune Estimation Resource, Human Protein Atlas, STRING, GeneMANIA, cBioPortal, and Linked Omics. Our investigation showed that CDKN2A was highly expressed in colon adenocarcinomas (P < .001). It is weakly expressed or not expressed in normal tissues. The survival time of patients with colon adenocarcinoma with high CDKN2A expression is significantly shorter than that of patients with low expression levels (P = .011). There was a significant positive correlation between the expression level of CDKN2A in colon adenocarcinoma tissues and the infiltration of CD4+ T cells, macrophages, and neutrophils. Moreover, there was a significant negative association between the expression level of CDKN2A in colon adenocarcinoma tissues and B cell infiltration. The ten hub genes included tumor protein 53, V-myc Avian Myelocytomatosis Viral Oncogene Homolog, AKT serine/threonine kinase 1, cyclin-dependent kinase 2, phosphatase and tensin homolog deleted on chromosome ten, cyclin D1, cyclin dependent kinase 4, cyclin dependent kinase inhibitor 1A, catenin beta 1, and B-Raf proto-oncogene, serine/threonine kinase. Mutations in the CDKN2A genome in colon adenocarcinoma reduce survival. Gene ontology and Kyoto Encyclopedia of Genes and Genomes analyses showed that the differentially expressed genes were enriched in apoptotic signaling pathways and multiple pathways related to metabolic progression. Our results indicate that CDKN2A can be used as a marker of poor prognosis in patients with colon adenocarcinoma. CDKN2A may regulate the occurrence and development of colon adenocarcinomas by influencing immune cell infiltration and metabolic pathways.

DeepKINET: a deep generative model for estimating single-cell RNA splicing and degradation rates

Messenger RNA splicing and degradation are critical for gene expression regulation, the abnormality of which leads to diseases. Previous methods for estimating kinetic rates have limitations, assuming uniform rates across cells. DeepKINET is a deep generative model that estimates splicing and degradation rates at single-cell resolution from scRNA-seq data. DeepKINET outperforms existing methods on simulated and metabolic labeling datasets. Applied to forebrain and breast cancer data, it identifies RNA-binding proteins responsible for kinetic rate diversity. DeepKINET also analyzes the effects of splicing factor mutations on target genes in erythroid lineage cells. DeepKINET effectively reveals cellular heterogeneity in post-transcriptional regulation.

The burden of liver cancer in Mongolia from 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019.

Liver cancer remains the leading cause of death and public health threat among the Mongolian population. So far, there has been no in-depth analysis to describe the burden of common attributable factors to liver cancer in Mongolia. Therefore, we aimed to explore the most prevalent causes of liver cancer and its trends from 1990 to 2019. We extracted the primary liver cancer data from the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2019 to examine the mortality and morbidity of liver cancer by its etiological types, which included alcohol, viral hepatitis B and C, and non-alcoholic steatohepatitis (NASH). The data was extracted by sex and 5-year age intervals from 1990 to 2019. Data included mortality, incidence, years of life lost (YLLs), years lived with disability (YLDs), and disability-adjusted life-years (DALYs) of liver cancer among the Mongolian population. Mongolia had the world's highest age-standardized DALYs for liver cancer (2558.1) in 2019. Alcohol-attributable DALYs (786.6) were 29 times higher than the global average (26.1), and liver cancer due to hepatitis C (752.6) and B (763.2) were 21.5 (35.0) and 10.9 (69.1) times higher, respectively. Over the past 30 years, there has been a steady increase in the incidence and number of deaths caused by liver cancer in Mongolia. In 2019, liver cancer incidence due to alcohol consumption was 3.1 times higher for males than females, and hepatitis B was 2.7 times higher for males than females. However, the incidence of hepatitis C and NASH were slightly higher for females. Deaths from liver cancer accounted for 9.51% (2365) of total deaths in Mongolia in 2019, with a continuously increasing trend in the fraction of death compared to 1990, which was 11 times higher than the global average (0.86%), particularly in females with a 319.6% (95% UI 234.9-435.7) increase observed during the study period. Liver cancer due to hepatitis B, C, and alcohol each shared about one-third of liver cancer deaths. A comprehensive analysis of the burden of liver cancer in Mongolia reveals alcohol use as a primary cause of liver cancer mortality, particularly affecting men and significantly impacting the disease burden. Viral hepatitis continues to pose a major public health concern in the country. Although significant milestones have progressed, addressing the unique demographic and geographical challenges requires tailored approaches for specific target populations. The evidence generated from this analysis is crucial to support policy guidance, contribute to evidence-based decisions, guide public health prevention measures, and amplify population health promotion and disease prevention throughout Mongolia.

Transcriptomic Landscape of Colorectal Mucinous Adenocarcinoma has Similarity with Intestinal Goblet Cell Differentiation

Introduction: Colorectal mucinous adenocarcinoma (MC) differs from adenocarcinoma (AD) in clinical features and molecular characteristics. The current treatment of colorectal MC is not precise enough, and the molecular characteristics remain unclear. The study aims to explore the difference between colorectal MC and AD on the transcriptome level for the possibility of treating colorectal MC precisely. Methods: The data of colorectal cancer (CRC) patients from The Cancer Genome Atlas (TCGA) database was assessed, and then differential analysis and weighted gene co-expression network analysis (WGCNA) were performed to identify the differential hub RNAs between colorectal MC and AD. Differential hub lncRNAs and hub RNA of significant modules were validated by quantitative real-time PCR (qRT-PCR) among different colon cancer cell lines. Results: In total, 1680 differential expressed RNAs (DERs) were found by comparing colorectal MC (52, 13.3%) with AD (340, 86.7%). Through the WGCNA, a mucin-associated RNA module was identified, while some others might be associated with unique immune progress. Finally, 6 differential hub RNAs in the mucin-associated RNA module (CTD-2589M5.4, RP11-234B24.2, RP11-25K19.1 and COLCA1) were validated by qRT-PCR and showed higher expression levels in mucin-producing colorectal cell lines (Ls174T and HT-29). Conclusion: This study suggests that clinical treatments for colorectal MC should be differentiated from AD. Further exploration of enterocyte (goblet cell) differentiation with tumor genesis and the distinct immune progression of MC may help to identify key therapeutic targets for colorectal MC. Further research on the application of immunotherapy to colorectal MC is needed.

Exploring time-to-event analysis models: A study of accelerated failure time and weighted least square estimation

Background: This study explores time-to-event analysis models, focusing on the comparison between accelerated failure time (AFT) and weighted least square estimation (WLSE) methods in analyzing survival data, emphasizing the challenge of heteroscedasticity in real-world datasets. Methods: Using the North Central Cancer Treatment Group (NCCTG) Lung Cancer dataset, five parametric AFT distributions (exponential, Weibull, log-logistic, log-normal, and generalized gamma) were evaluated. The log-likelihood method was utilized to calculate the AIC and BIC values for each distribution to determine the best fit model. Results: Among the different distributions tested, the Weibull distribution was identified as the best fit model based with the lowest AIC and BIC values of 2311.70 and 2318.56, respectively. This distribution played a crucial role in determining factors associated with the cancer data. Specifically, the variablecalories consumed at a meal (meal.cal) was associated with cancer patient recovery, with an estimate of 0.00004 and a standard error of 0.0002. Results indicated homoscedastic data through the Breusch-Pagan test. Moreover, the results demonstrated that when dealing with homoscedastic data, the Weibull AFT model outperformed WLSE in producing statistically significant effects and precise estimates. Conclusion: For data exhibiting homoscedasticity, the AFT method yielded superior results compared to WLSE in terms of statistical significance and precision. Therefore, it is advisable to employ AFT for homoscedastic data and WLSE for heteroscedastic data to enhance the accuracy of covariate effect estimates in survival analysis.

Hybrid embedded and filter feature selection methods in big-dimension mammary cancer and prostatic cancer data

&lt;p&gt;The feature selection method enhances machine learning performance by enhancing learning precision. Determining the optimal feature selection method for a given machine learning task involving big-dimension data is crucial. Therefore, the purpose of this study is to make a comparison of feature selection methods highlighting several filters (information gain, chi-square, ReliefF) and embedded (Lasso, Ridge) hybrid with logistic regression (LR). A sample size of n=100, 75 is chosen randomly, and the reduction features d=50, 22, and 10 are applied. The procedure for feature reduction makes use of the entire sample sizes. Each sample size's results are compared, including tests with no feature selection process. The results indicate that LR+ReliefF is the best method for mammary cancer data, whereas LR+IG is the best for prostatic cancer data, making the filter more suitable than embedded for big-dimension data. This study revealed that the sample's features and size influence the most effective method for selecting features from big-dimension data. Therefore, it provides insight into the most effective methods for particular features and sample sizes in high-dimensional data.&lt;/p&gt;

Federated unsupervised random forest for privacy-preserving patient stratification.

In the realm of precision medicine, effective patient stratification and disease subtyping demand innovative methodologies tailored for multi-omics data. Clustering techniques applied to multi-omics data have become instrumental in identifying distinct subgroups of patients, enabling a finer-grained understanding of disease variability. Meanwhile, clinical datasets are often small and must be aggregated from multiple hospitals. Online data sharing, however, is seen as a significant challenge due to privacy concerns, potentially impeding big data's role in medical advancements using machine learning. This work establishes a powerful framework for advancing precision medicine through unsupervised random forest-based clustering in combination with federated computing. We introduce a novel multi-omics clustering approach utilizing unsupervised random forests. The unsupervised nature of the random forest enables the determination of cluster-specific feature importance, unraveling key molecular contributors to distinct patient groups. Our methodology is designed for federated execution, a crucial aspect in the medical domain where privacy concerns are paramount. We have validated our approach on machine learning benchmark datasets as well as on cancer data from The Cancer Genome Atlas. Our method is competitive with the state-of-the-art in terms of disease subtyping, but at the same time substantially improves the cluster interpretability. Experiments indicate that local clustering performance can be improved through federated computing. The proposed methods are available as an R-package (https://github.com/pievos101/uRF).

Association of overexpression of PLD6, CHRAC1 and PDCD5 with type 2 diabetes mellitus

&amp;lt;p&amp;gt;&amp;lt;strong&amp;gt;Aim &amp;lt;/strong&amp;gt;Diabetes type 2 (DT2) is a metabolic disease characterized by high blood sugar caused by insulin resistance and/or insufficient insulin production. The pathogenesis of DT2 is complicated by both genetic predisposition and environmental and lifestyle variables. At least 150 genetic variants have been linked to the probability of having DT2. The aim of this study was to determine the expression of PLD6, CHRAC1, and PDCD5 genes in type 2 diabetic patients.&amp;amp;nbsp;&amp;lt;br /&amp;gt;&amp;lt;strong&amp;gt;Methods &amp;lt;/strong&amp;gt;Information on 12 DT2 patients was obtained from the Gene Expression Omnibus (GEO) using the series identification (ID) (GSE34008). The analysis tools GEO2R, String Utils (STRING), University of Alabama at Birmingham Cancer data analysis (UAL-CAN), and the Cancer Genome Atlas (TCGA) were used. The human protein atlas provided details on gene cancer.&amp;lt;br /&amp;gt;&amp;lt;strong&amp;gt;Results &amp;lt;/strong&amp;gt;Only ten genes with expression differences ranging from low to high were selected. PLD6, CHRAC1, and PDCD5 were detected to have higher expression in patients compared to controls. The number of patients with primary pancreatic adenocarcinoma for SLC16A4, DERK2, and CHRAC1 was greater than that of healthy controls. Concerning the severity of cancer, all chosen genes demonstrated a greater proportion of affected individuals compared to the control group.&amp;amp;nbsp;&amp;lt;br /&amp;gt;&amp;lt;strong&amp;gt;Conclusion &amp;lt;/strong&amp;gt;There are multiple genes whose increased expression is linked to type 2 diabetes.&amp;lt;/p&amp;gt;

Expression and prognostic impact of VDAC3 in colorectal adenocarcinoma.

Colorectal adenocarcinoma (COAD) is a malignant tumor with high mortality and low 5-year survival rate. Voltage-dependent anion channel 3 (VDAC3) is the least understood isoform of voltage-dependent anion-selective channels in the mitochondrial outer membrane. In this thesis, we aimed to investigate the prognostic value of VDAC3 and provide new insights into colon adenocarcinoma. We utilized The Cancer Genome Atlas (TCGA) database, Gene Expression Omnibus (GEO) database, Human Protein Atlas online database, and the University of ALabama at Birmingham CANcer data analysis Portal (UALCAN) database to analyze VDAC3 expression in COAD and assess patient survival rates. Univariate and multivariate Cox regression analyses were employed to evaluate VDAC3's prognostic significance for COAD. Gene set variation analysis (GSVA) was utilized to explore COAD-related signaling pathways associated with VDAC3. Additionally, we predicted the relationship between VDAC3 expression and anticancer drug sensitivity using the CellMiner database. In the TCGA database, VDAC3 demonstrated elevated expression levels in COAD, which was further validated by findings from the GEO database. Survival analysis conducted using Kaplan-Meier (K-M) curves highlighted that the patients with decreased VDAC3 expression exhibited significantly shorter overall survival durations. VDAC3 expression demonstrated correlation with COAD pathological stage. VDAC3 gene mutation was linked to COAD outcomes. Cox regression analysis showed that VDAC3 was an independent predictor. In addition, GSVA analysis showed that VDAC3 was closely related to mitochondria-related biological processes and involved in the occurrence and development of mitochondria-related diseases. Finally, analysis of the CellMiner database predicted that VDAC3 expression was positively correlated with chelerythrine and cladribine, but negatively correlated with Ergenyl. Our study suggests that VDAC3 may be a potential biomarker for early diagnosis, prognosis, and treatment of COAD.

Recovering single-cell expression profiles from spatial transcriptomics with scResolve.

Many popular spatial transcriptomics techniques lack single-cell resolution. Instead, these methods measure the collective gene expression for each location from a mixture of cells, potentially containing multiple cell types. Here, we developed scResolve, a method for recovering single-cell expression profiles from spatial transcriptomics measurements at multi-cellular resolution. scResolve accurately restores expression profiles of individual cells at their locations, which is unattainable with cell type deconvolution. Applications of scResolve on human breast cancer data and human lung disease data demonstrate that scResolve enables cell-type-specific differential gene expression analysis between different tissue contexts and accurate identification of rare cell populations. The spatially resolved cellular-level expression profiles obtained throughscResolve facilitate more flexible and precise spatial analysis that complements raw multi-cellular level analysis.

SEMA7A as a Novel Prognostic Biomarker and Its Correlation with Immune Infiltrates in Breast Cancer.

The role of Semaphorin 7a (SEMA7A) in the initiation and progression of different types of cancerous lesions has been extensively studied. However, the prognostic significance of SEMA7A, specifically in breast cancer (BC), lacks clarity. We conducted an evaluation on the relationship between SEMA7A and the prognosis, immune invasion and tumor mutation burden in different types of cancer by analyzing data from The Cancer Genome Atlas database. The present study focused on investigating the expression level, mutation, immune correlation and coexpression of SEMA7A in BC, utilizing various databases such as the University of Alabama at Birmingham Cancer data analysis portal, cBioPortal and tumor immune estimation resource. Survival analysis was carried out using the Kaplan-Meier Plotter. Furthermore, we employed the R software package to generate receiver operating characteristic (ROC) curves and nomograms. Notably, P<0.05 was considered to indicate statistical significance. Using pancancer analysis, it has been observed that the expression of SEMA7A is elevated in various types of cancer and is strongly correlated with the prognosis of different cancer types. SEMA7A also exhibits a significant association with the tumor mutation burden of diverse types of cancer. Moreover, SEMA7A displays a notable increase in BC cases, and was indicated to have a substantial association with the abundance of immune infiltration. In-depth survival analysis demonstrated that elevated levels of SEMA7A expression are notably linked to shorter overall survival and distant metastasis-free survival among patients with BC. The efficiency of SEMA7A as a reliable prognostic biomarker for BC has been substantiated by the validation of ROC curves and nomograms. SEMA7A has the potential to function as a prognostic indicator for BC, and its correlation with immune infiltration in BC is significant.

Predicting spatially resolved gene expression via tissue morphology using adaptive spatial GNNs.

Spatial transcriptomics technologies, which generate a spatial map of gene activity, can deepen the understanding of tissue architecture and its molecular underpinnings in health and disease. However, the high cost makes these technologies difficult to use in practice. Histological images co-registered with targeted tissues are more affordable and routinely generated in many research and clinical studies. Hence, predicting spatial gene expression from the morphological clues embedded in tissue histological images provides a scalable alternative approach to decoding tissue complexity. Here, we present a graph neural network based framework to predict the spatial expression of highly expressed genes from tissue histological images. Extensive experiments on two separate breast cancer data cohorts demonstrate that our method improves the prediction performance compared to the state-of-the-art, and that our model can be used to better delineate spatial domains of biological interest. https://github.com/song0309/asGNN/.

CIBRA identifies genomic alterations with a system-wide impact on tumor biology.

Genomic instability is a hallmark of cancer, leading to many somatic alterations. Identifying which alterations have a system-wide impact is a challenging task. Nevertheless, this is an essential first step for prioritizing potential biomarkers. We developed CIBRA (Computational Identification of Biologically Relevant Alterations), a method that determines the system-wide impact of genomic alterations on tumor biology by integrating two distinct omics data types: one indicating genomic alterations (e.g. genomics), and another defining a system-wide expression response (e.g. transcriptomics). CIBRA was evaluated with genome-wide screens in 33 cancer types using primary and metastatic cancer data from the Cancer Genome Atlas and Hartwig Medical Foundation. We demonstrate the capability of CIBRA by successfully confirming the impact of point mutations in experimentally validated oncogenes and tumor suppressor genes (0.79 AUC). Surprisingly, many genes affected by structural variants were identified to have a strong system-wide impact (30.3%), suggesting that their role in cancer development has thus far been largely under-reported. Additionally, CIBRA can identify impact with only 10 cases and controls, providing a novel way to prioritize genomic alterations with a prominent role in cancer biology. Our findings demonstrate that CIBRA can identify cancer drivers by combining genomics and transcriptomics data. Moreover, our work shows an unexpected substantial system-wide impact of structural variants in cancer. Hence, CIBRA has the potential to preselect and refine current definitions of genomic alterations to derive more nuanced biomarkers for diagnostics, disease progression, and treatment response. The R package CIBRA is available at https://github.com/AIT4LIFE-UU/CIBRA.

Bayesian Optimal Two-Sample Tests for High-Dimensional Gaussian Populations

We propose minimax optimal Bayesian two-sample tests for testing equality of high-dimensional mean vectors and covariance matrices between two populations. In many applications including genomics and medical imaging, it is natural to assume that only a few entries of two mean vectors or covariance matrices are different. Many existing tests that rely on aggregating the difference between empirical means or covariance matrices are not optimal or yield low power under such setups. Motivated by this, we develop Bayesian two-sample tests employing a divide-and-conquer idea, which is powerful especially when the differences between two populations are rare but large. The proposed two-sample tests manifest closed forms of Bayes factors and allow scalable computations even in high-dimensions. We prove that the proposed tests are consistent under relatively mild conditions compared to existing tests in the literature. Furthermore, the testable regions from the proposed tests turn out to be minimax optimal in terms of rates. Simulation studies show clear advantages of the proposed tests over other state-of-the-art methods in various scenarios. Our tests are also applied to the analysis of the gene expression data of two cancer data sets.