Integration of multiomics features for blood-based early detection of colorectal cancer.

BACKGROUND: Early detection of colorectal cancer (CRC) significantly enhances patient outcomes. Conventional CRC screening tools, like endoscopy and stool-based tests, have constraints due to their invasiveness or suboptimal patient adherence. Recently, liquid biopsy employing plasma cell-free DNA (cfDNA) has emerged as a potential noninvasive screening technique for various malignancies. METHODS: In this research, we harnessed the Mutation Capsule Plus (MCP) technology to profile an array of genomic characteristics from cfDNA procured from a single blood draw. This profiling encompassed DNA methylation, the 5' end motif, copy number variation (CNV), and genetic mutations. An integrated model built upon selected multiomics biomarkers was trained using a cohort of 93 CRC patients and 96 healthy controls. RESULTS: This model was subsequently validated in another cohort comprising 89 CRC patients and 95 healthy controls. Remarkably, the model achieved an area under the curve (AUC) of 0.981 (95% confidence interval (CI), 0.965-0.998) in the validation set, boasting a sensitivity of 92.1% (95% CI, 84.5%-96.8%) and a specificity of 94.7% (95% CI, 88.1%-98.3%). These numbers surpassed the performance of any single genomic feature. Importantly, the sensitivities reached 80% for stage I, 89.2% for stage II, and were 100% for stages III and IV. CONCLUSION: Our findings underscore the clinical potential of our multiomics liquid biopsy test, indicating its prospective role as a noninvasive method for early-stage CRC detection. This multiomics approach holds promise for further refinement and broader clinical application.

Association of XRCC2 with breast cancer, a multi-omics analysis at genomic, transcriptomic, and epigenomic level.

One of the main causes of cancer-related mortality for women worldwide is breast cancer (BC). The XRCC2 gene, essential for DNA repair, has been implicated in cancer susceptibility. This study aims to evaluate the association between XRCC2 and BC risk. The study was conducted at Zheen International Hospital in Erbil, Iraq, between 2021 and 2024 with a total of 88 samples, including 44 paired normal and cancer tissue samples. Mutation analysis was performed using Next-Generation Sequencing, coupled with in silico tools for variant impact prediction. Expression levels were assessed through RT-PCR, and methylation status was determined using methylation-sensitive restriction enzyme digestion PCR. The study identified seven inherited germline variants in the XRCC2 gene, with five of these mutations being Uncertain Significance, one being Likely Pathogenic, and one being Likely benign. RNA purity was found high with mean A260/280 ratios of 1.986 ± 0.097 in normal (N) and 1.963 ± 0.092 in tumor (T) samples. Tumor samples exhibited a higher RNA concentration (78.56 ± 40.87 ng/µL) than normal samples (71.44 ± 40.79 ng/µL). XRCC2 gene expression was significantly upregulated in tumor tissue, with marked increases in patients aged 40-55 and >56 years and in higher cancer grades (II and III) and invasive ductal carcinoma (p-values ranging from <0.0001 to 0.0392). DNA methylation rates in tumor tissues were low (7%), suggesting limited regulation by methylation. The study suggests that XRCC2 can be classified as an oncogene and that its structural investigation by targeted NGS and expression evaluation can be used as a potential biomarker in BC.

A model-free and distribution-free multi-omics integration approach for detecting novel lung adenocarcinoma genes.

Detection of important genes affecting lung adenocarcinoma (LUAD) is critical to finding effective therapeutic targets for this highly lethal cancer. However, many existing approaches have focused on single outcomes or phenotypic associations, which may not be as thorough as investigating molecular transcript levels within cells. In this article, we apply a novel multivariate rank-distance correlation-based gene selection procedure (MrDcGene) to LUAD multi-omics data downloaded from The Cancer Genome Atlas (TCGA). MrDcGene provides additional opportunities for detecting novel susceptibility genes as it leverages information from multiple platforms, while efficiently handling challenges such as high dimensionality, low signal-to-noise ratio, unknown distributions, and non-linear structures, etc. Notably, the MrDcGene method is able to detect two different scenarios, i.e., strong association strength with a few gene expressions and weak association strength with several gene expressions. After thoroughly exploring the association between gene expression (GE) and multiple other platforms, including reverse phase protein array (RPPA), miRNA, copy number variation (CNV) and DNA methylation (ME), we detect several novel genes that may play an important role in LUAD (ZNF133, CCDC159, YWHAZ, HNRNPR. ITPR2, PTHLH, and WIPI2). In addition, we quantitatively validate several other susceptibility genes that were reported in the literature using different methods and studies. The accuracy of the MrDcGene approach is theoretically assured and empirically demonstrated by the simulation studies.

The prediction of drug sensitivity by multi-omics fusion reveals the heterogeneity of drug response in pan-cancer.

Cancer drug response prediction based on genomic information plays a crucial role in modern pharmacogenomics, enabling individualized therapy. Given the expensive and complexity of biological experiments, computational methods serve as effective tools in predicting cancer drug sensitivity. In this study, we proposed a novel method called Multi-Omics Integrated Collective Variational Autoencoders (MOICVAE), which leverages integrated omics knowledge, including genomic and transcriptomic data, to fill in missing cancer-drug associations and enhance drug sensitivity prediction. Our method employs an encoder-decoder network to learn latent feature representations from cell lines. These learned feature vectors are then fed into a collective variational autoencoder network to train an association matrix. We evaluated MOICVAE on the GDSC and CCLE benchmark datasets using 10-fold cross-validation and achieved impressive AUCs of 0.856 and 0.808, respectively, outperforming state-of-the-art methods. Furthermore, on the TCGA dataset, consisting of 25 drugs across 7 cancer types, MOICVAE exhibited an average AUC of 0.91 in predicting drug sensitivity. Additionally, significant differences were observed in survival, tumor inflammatory assessment, and tumor microenvironment between the predicted drug-sensitive and drug-resistant groups. These results are consistent with predictions made on the METABRIC dataset. Moreover, we discovered that fusing omics data based on mRNA and CNV (copy number variations) yielded superior results in drug sensitivity prediction. MOICVAE not only achieved higher accuracy in drug sensitivity prediction but also provided additional value for combining immunotherapy with chemotherapy, offering patients with more precise treatment options. The code and dataset for MOICVAE are freely available at https://github.com/wanggnoc/MOICVAE.

Geometric graph neural networks on multi-omics data to predict cancer survival outcomes.

The advance of sequencing technologies has enabled a thorough molecular characterization of the genome in human cancers. To improve patient prognosis predictions and subsequent treatment strategies, it is imperative to develop advanced computational methods to analyze large-scale, high-dimensional genomic data. However, traditional machine learning methods face a challenge in handling the high-dimensional, low-sample size problem that is shown in most genomic data sets. To address this, our group has developed geometric network analysis techniques on multi-omics data in connection with prior biological knowledge derived from protein-protein interactions (PPIs) or pathways. Geometric features obtained from the genomic network, such as Ollivier-Ricci curvature and the invariant measure of the associated Markov chain, have been shown to be predictive of survival outcomes in various cancers. In this study, we propose a novel supervised deep learning method called geometric graph neural network (GGNN) that incorporates such geometric features into deep learning for enhanced predictive power and interpretability. More specifically, we utilize a state-of-the-art graph neural network with sparse connections between the hidden layers based on known biology of the PPI network and pathway information. Geometric features along with multi-omics data are then incorporated into the corresponding layers. The proposed approach utilizes a local-global principle in such a manner that highly predictive features are selected at the front layers and fed directly to the last layer for multivariable Cox proportional-hazards regression modeling. The method was applied to multi-omics data from the CoMMpass study of multiple myeloma and ten major cancers in The Cancer Genome Atlas (TCGA). In most experiments, our method showed superior predictive performance compared to other alternative methods.

MOPA: An integrative multi-omics pathway analysis method for measuring omics activity.

Pathways are composed of proteins forming a network to represent specific biological mechanisms and are often used to measure enrichment scores based on a list of genes in means to measure their biological activity. The pathway analysis is a de facto standard downstream analysis procedure in most genomic and transcriptomic studies. Here, we present MOPA (Multi-Omics Pathway Analysis), which is a multi-omics integrative method that scores individual pathways in a sample wise manner in terms of enriched multi-omics regulatory activity, which we refer to mES (multi-omics Enrichment Score). The mES score reflects the strength of regulatory relations between multi-omics in units of pathways. In addition, MOPA is able to measure how much each omics contribute to mES that may be used to observe what kind of omics are active in a pathway within a sample group (e.g., subtype, gender), which we refer to OCR (Omics Contribution Rate). Using nine different cancer types, 93 clinical features and three types of omics (i.e., gene expression, miRNA and methylation), MOPA was used to search for clinical features that were explainable in context of multi-omics. By evaluating the performance of MOPA, we showed that it yielded higher or at least equal performance compared to previous single and multi-omics pathway analysis tools. We find that the advantage of MOPA is the ability to explain pathways in terms of omics relation using mES and OCR. As one of the results, the TGF-beta signaling pathway was captured as an important pathway that showed distinct mES and OCR values specific to the CMS4 subtype in colon adenocarcinoma. The mES and OCR metrics suggested that the mRNA and miRNA expressions were significantly different from the other subtypes, which was concordant with previous studies. The MOPA software is available at https://github.com/jaeminjj/MOPA.