A review on trends in development and translation of omics signatures in cancer

Abstract

The field of cancer genomics and transcriptomics has evolved from targeted profiling to swift sequencing of individual tumor genome and transcriptome. The steady growth in genome, epigenome, and transcriptome datasets on a genome-wide scale has significantly increased our capability in capturing signatures that represent both the intrinsic and extrinsic biological features of tumors. These biological differences can help in precise molecular subtyping of cancer, predicting tumor progression, metastatic potential, and resistance to therapeutic agents. In this review, we summarized the current development of genomic, methylomic, transcriptomic, proteomic and metabolic signatures in the field of cancer research and highlighted their potentials in clinical applications to improve diagnosis, prognosis, and treatment decision in cancer patients.

Keywords: Cancer genomics; DNA methylation; Machine learning; Mutational signatures; Transcriptomics; Translational Medicine.

Fig. 1

Timeline of milestones of different omics signatures.Fig. 1. The figure shows some of the key milestones for 1. Mutational signatures, 2. Methylation signature, 3. Transcriptome signature and 4. Proteomic and metabolic signatures. Abbreviation: SNV, single-nucleotide variant; SV, structural variation; CNV, copy number variation; SBS, single base substitution; DBS, double base substitution; ID, insertions and deletion; TMZ, temozolomide; DKFZ, The German Cancer Research Center; AML, acute myeloid leukemia; ALL, acute lymphocytic leukemia; NSCLC, non-small cell lung cancer.

Fig. 2

Summary of mutational signatures based on SNV, CNV and SV.Fig. 2. The figure depicts features of SNV, SV and CNV mutational signatures. For SNVs, the 5’ and 3’ of the mutated site, as well as the type of mutation are considered. For SVs, four types of variants, the distance between the two breakpoints, along with the clustered status of the SVs are considered. For CNVs, factors such as heterozygosity, total copy number, and CNV size are taken into consideration. Mutational signature fitting is a mathematical procedure used to determine the combination of known signatures, such as the COSMIC catalogue. Abbreviation: SNV, single-nucleotide variant; SV, structural variation; CNV, copy number variation; HD, homozygous deletion; LOH, loss of heterozygosity; Het, heterozygous.

Fig. 3

Summary of general workflow of constructing the classification model from DNA methylation data.Fig. 3. The figure depicts the basic workflow of construction of DNA methylation classifier. Both tumor and cell-free DNA can be used as input. The methylation level of CpGs is measured either using targeted methylation microarray or bisulfite sequencing if genome-wide data is needed. CpGs significantly associated with the feature of interest (e.g. tumor subtype) are selected using models such as MLR. The clustering result is checked using tools like t-SNE or UMAP. Finally, the selected probes are submitted to supervised classifier-building algorithms such as SVM and RF. Abbreviation: MR, multivariate regression; SVM, support vector machine; RF, random forest.

Fig. 4

Examples of clustering algorithms in analyzing transcriptome signature.Fig. 4. The figure shows examples of unsupervised clustering algorithms used in identifying molecular subtypes of cancers using transcriptomic data. Network-based algorithms such as spectral clustering, MCL and Louvain identifies clusters without making prior assumptions about the nature of clusters in the data. Traditional ML methods start with selection of subtype associated DEGs using regression models, followed by submitting them into ML algorithms like consensus clustering, SVM and RF. Deep learning models gain increased interests in recent years. Whole-transcriptome data can be used directly in training models like CNN and GCNN without the need for prior filtering of DEGs. Abbreviation: CNN, convolutional neural network; DEG, differentially expressed gene; GCNN, graph convolutional neural networks; LASSO, least absolute shrinkage and selection operator; MCL, Markov cluster algorithm; ML, machine learning; MR, multivariate regression; SVM, support vector machine; RF, random forest.