1. Introduction
Globally, colorectal cancer (CRC) is the second leading cause of cancer death in both males and females. In 2020, there were an estimated 935,000 mortalities from CRC, which accounted for over 1.9 million cases [
1]. CRC has become a major public health concern in the Asia-Pacific region, including Malaysia [
2]. Malaysia is a developing country located in Southeast Asia with an estimated population of 32.7 million in 2021 and an annual population growth rate of 0.2%. The Malaysian National Cancer Registry Report 2012–2016 found that CRC is the most common cancer in males (16.9% of all cancers diagnosed) and the second most common cancer in females (10.7% of all cancers diagnosed) [
3]. A study of the National Cancer Registry for CRC from 2008 to 2013 revealed that the overall age-standardized incidence rate for CRC was 21.32 per 100,000 population. Those of Chinese ethnicity had the highest CRC incidence rate (27.35), followed by Malay (18.95) and Indian (17.55) [
4]. Despite the severity caused by CRC, local studies pertaining to this field are scarce. The first study on the local CRC omics has been conducted on 50 CRC patients from Hospital Canselor Tunku Muhriz, Cheras, Kuala Lumpur. Data from the 50 samples was used as a model for CRC database development, and it will also guide the next experiments on what type of data/samples will be needed to ensure the development of a comprehensive repository. In addition, TCGA-My was developed with the aim of providing a repository platform for a Malaysian cancer consortium. Cancer databases have been developed and curated using bioinformatics with the support of highly advanced experimental evidence to assist the discovery of novel and unknown information about cancer.
Even though there is a growing trend in CRC incidence and mortality, the pathobiology of CRC remains unknown [
1,
3], hence the need for a CRC omics database that will provide pools of genes, proteins or metabolites that are associated with this cancer, which can be identified through integrative analysis. To this end, genomic and metabolomic data was collected from 50 Malaysian CRC patients, and a database was developed to provide a systematic data storage and retrieval platform, which is known as Malaysian TCGA (TCGA-My). TCGA-My stores CRC data for the Malaysian ethnic population, and this feature distinguishes TCGA-My from other colorectal cancer databases, such as Colorectal Cancer Atlas (
http://www.colonatlas.org (accessed on 10 May 2022)) [
5], Colorectal Cancer Biomarker Database (CBD) (
http://sysbio.suda.edu.cn/CBD/ (accessed on 10 May 2022)) [
6] and Colon Rectal Cancer Gene Database (CoReCG) (
https://lms.snu.edu.in/corecg/ (accessed on 10 May 2022)) [
7], which store the genes or proteins associated with CRC regardless of the patients’ region of origin.
In TCGA-My, most of the genes and metabolites are fully annotated to help users explore functional information for the data. Biological information for chromosomes, DNA region, tissue occurrence and type of mutation for each variant are also included. Circos plots are generated and provided to help users visualize the variants on human chromosomes. Additionally, details for the samples such as gender, age and ethnicity are also provided as they are useful for the interpretation of various analyses of the data. Eventually, TCGA-My will provide CRC raw data and results from the integrative analysis that can be used to predict phenotypic changes in cancer cells upon chemo-immunotherapy treatment toward precision and personalized medicine initiatives. With these features, TCGA-My should serve as a comprehensive database that will be an excellent aid in providing and integrating accurate molecular information to understand the relationship between CRC and ethnicity.
We envision TCGA-My will expand and serve as a Malaysian cancer multi-omics data repository and provide a unique opportunity for a systems biology approach to tackle the complexity of cancer cells and CRC pathobiology through the unification of experimental data and computational/mathematical models. Currently, TCGA-My stores 1,517,841 variants, 23,695 genes and 167,451 metabolites obtained from 13 genomic and 50 metabolomic samples. Genomic experiments with another 37 samples are in progress and the data will be gradually deposited into this database. We used the available data to design the TCGA-My architecture. The development of TCGA-My is ongoing and accessible at
https://tcgamy.inbiosis.org/(TCGA-My v1.0, last updated April 2022) (accessed on 10 May 2022).
3. Results
3.1. Database Summary
Figure 2 illustrates the organization of four datasets, i.e., sample; variants with an ID obtained from COSMIC and dbSNP databases; genes with an ID retrieved from PDB, RefSeq ncRNA and UniProt; and metabolites.
Table 1 summarizes the number of entries for each dataset.
3.2. Main Datasets
3.2.1. Samples
Each sample of CRC represents data from different patients. Each sample was provided with the patients’ details, such as gender, age, ethnicity, diagnosis information, anatomical location and cancer stage (
Table 2). Two cancer staging systems were included in this database, i.e., TNM [
29] and Dukes staging [
30] (
Table 3).
3.2.2. Variants
TCGA-My provides a list of variants with comprehensive mutation details that includes chromosomal location, variant start and end point, reference and alternate nucleotide, variant position in DNA region, affected gene, tissue occurrence and type of mutation caused by the variant.
Figure 3 shows the correlation between the number of variants, gender and age of the CRC patients. The number of variants in 12 samples showed slight differences with a range of 95,000± to 116,000± variants. Only one sample (C474), a female with the oldest age of 79, showed the highest number of variants (751,000± variants). Age might influence the number of variants [
31].
Table 4 and
Table 5 describe the position of variants in DNA regions and mutations in genome data, respectively, for 13 CRC patients. Circos plots display the position of the variants on 24 human chromosomes and the mutation types in a sample. A Circos plot was constructed for every patient and
Figure 4 shows an example of the plot constructed for patient C474.
3.2.3. Genes
The variant genes were provided with gene symbol descriptions and linked with three different IDs from ncRNA, PDB and UniProt IDs. Forty-four genes were recently annotated as CRC driver genes by Abdullah and Muhammad [
16], where a computational approach was applied to identify CRC driver genes using a bioinformatics pipeline that consisted of the Cancer Genome Interpreter (CGI) [
32] and Integrated Cancer Genome Score (iCAGES) [
33] analysis platforms. Driver genes refer to genes whose mutations promote tumor growth [
34], and further investigation of these genes is critical in precision oncology [
35].
Table 6 lists the number of genes and driver genes for each patient.
3.2.4. Metabolites
The metabolites in TCGA-My were described with the name, class and mass. From 89,256 identified metabolites, only eleven metabolites were significantly altered, suggesting their potential as biomarkers for CRC (
Figure 5).
3.3. Database Interface and Access
The TCGA-My interface contains four main menus, i.e., About, Browse, Search and Help, that were designed to help users easily navigate the respective pages. Each primary page has a header with a logo, search box and tabs that can be used to navigate to the primary pages and the dropdown menu. The logo on the header can be clicked to return to the homepage and the search box serves as the homepage search box.
Homepage displays statistics for the main datasets, i.e., sample, variant, gene and metabolites. The number of entries can be clicked, directing users to the respective dataset. Menus for about, browse and search functions were also included on this page. The search box on the homepage can be used to search any ID or terms in the datasets.
About page provides general information for TCGA-My and CRC.
Browse page lists all TCGA-My datasets, four of which are the main datasets and two datasets that contain functional annotation information. These datasets can also be retrieved from the dropdown tab, named Menu, which can be found at the header of each primary page. The datasets on this page are described as follows:
Sample dataset: contains genome and metabolome data for CRC. The sample datasets are categorized into gender and ethnicity.
Variant dataset: contains variants that were obtained from the 13 samples of genome data for CRC. This dataset is also categorized into chromosomes, DNA regions, tissue occurrence and type of mutations.
Gene dataset: contains genes affected by the variants. A list of driver genes can also be obtained.
Metabolite dataset: contains metabolites that are profiled in the metabolome data for CRC. This dataset is categorized into class and regulation.
Gene ontology dataset: contains GOs information (biological process, molecular function and cellular component) for variant genes in CRC.
Pathway dataset: contains pathways information for genes and metabolites of CRC.
Search page allows two search options, i.e., simple search and variant advanced search. Simple search serves a similar function to the search box on the homepage and the header of the primary pages. Advanced search allows the users to find variant(s) with a combination of different keywords. Users can conduct a quick search for the variants from a certain sample that are linked to a specific driver gene.
Help page provides an entity-relationship diagram and the table information deposited in this database. The entity-relationship diagram shows the relationship between datasets stored in this database. Table information defines all terms used in TCGA-My. Additionally, this page also provides the contact details for questions or invitations to collaborate.
Each entry in TCGA-My provides a brief description. For instance, if a user searches for one of the patient ID, for example, C187, it will redirect the user to the description page for C187. Four tabs for the mutations/variants, Circos, genes and metabolites identified from C187 data will appear. If a user clicks on one of the entries in the gene tab, for example, gene of ACACA, a description of ACACA will appear and display five tabs for sample, GOs, variants, KEGG and PANTHER pathways. All datasets in TCGA-My are freely downloadable in comma-separated values (CSV) format by clicking the “Download CSV” button located at the top-right of each table.
4. Discussion
4.1. Strength of TCGA-My
TCGA-My is the first database that houses genome and metabolome data for Malaysian CRC patients. Given the high number of patients and deaths caused by CRC in Malaysia [
1], this effort is vital for collecting and collating omics data for in-depth studies that will lead to the development of precision medicine in Malaysia. Details for each patient, such as gender, age, ethnicity, diagnosis information, anatomical location and cancer stage are provided along with other important information to facilitate a reliable prognosis, accurate diagnosis and effective treatment of cancer [
36,
37].
TCGA-My also contains lists of variants, related genes and metabolites, which are useful for researchers investigating and characterizing the molecules responsible for CRC in specific conditions, such as gender, age, ethnicity, anatomical location or cancer stage. In addition, the decision to treat CRC and other cancers relies on many factors, and each patient has a different condition, so one treatment is ineffective for all conditions. Precision medicine or personalized medicine benefits from omics data and it can provide customized treatment based on the patient’s molecular signatures that will be most effective. Such molecular datasets are crucial for finding potential biomarkers and signatures for CRC that can be used to design the most effective therapy and toward precision medicine [
38].
TCGA-My is an integrated functional database that contains relevant biological information for genomics and metabolomics data for CRC with links to external databases to ease users into exploring the genes and metabolites. For example, users can find further information about a particular gene by clicking on the GeneCards or UniProt symbol. Further information about GOs and pathways can also be found by clicking on the identifier for the Gene Ontology Consortium [
21], KEGG [
17] and PANTHER [
27] databases, and a window will automatically appear in your browser. In this way, the user is redirected from TCGA-My to a particular database and avoids manual browsing of the original database when searching for a gene of interest.
This database is systematically organized and provides a dynamic web interface. The users can find all information for each entry on a single page in a separate Table For example, the sample entry provides available information about the patients. All variants, related genes and metabolites can be found on a similar page in different tabs. The genomic data is displayed using a Circos plot to help visualization of the mutation that occurred in a particular genome. Related information on a gene entry can also be found on a similar page with different tabs.
TCGA-My was developed with the aim of providing a platform for a Malaysian cancer consortium, where a CRC dataset was used to provide a model for CRC database development that will be extended gradually. It will also guide the choice of data/samples that will be needed to ensure a comprehensive repository and provide accessible datasets for the epidemiological study of CRC.
4.2. Weaknesses of TCGA-My and Future Perspectives
One of the weaknesses of TCGA-My is the absence of another two omics datasets, which are transcriptomics and proteomics. Integrating all omics data for CRC, i.e., genomics, transcriptomics, proteomics and metabolomics, will provide clearer insights into the molecular mechanisms of CRC [
39]. Currently, TCGA-My only provides datasets for two major ethnicities in Malaysia, Malay and Chinese, and none for Indians and other minor ethnicities in Malaysia. Nevertheless, the Chinese have been shown to have the highest CRC incidence rate in Malaysia, followed by the Malays [
4]. Additionally, there are no datasets available for a younger generation with CRC. TCGA-My provides datasets for ages 41–82 years, but according to the descriptive statistics from the Malaysia Cancer Patient Registry-Colorectal Cancer, CRC incidence rises with age [
4]. In addition, early onset of CRC occurs at an age of more than 50 years old in many countries, such as the United States, Canada and Australia [
1]. Finally, the Circos plots in TCGA-My are static, making it difficult for users trying to identify the location of particular mutations that occurred in a sample.
Hence, it will be very important to periodically update the datasets in TCGA-My and further improve the interface in order to provide a comprehensive and convenient resource for other researchers to utilize.
4.3. Example of Applications
The first example concerns a molecular cancer researcher who wants to search for a specific gene of interest. Starting with the browse menu, a list of genes with known variants can be obtained, and the relevant information, such as the functional annotation of the molecule and patient statistics, can be retrieved. The users can download the datasets for more advanced analysis as per their intention.
The second example concerns bioinformaticians or biostatisticians interested in conducting big data analysis. With 1.5 million variants that can be downloaded from TCGA-My, there are several analyses that can be conducted, such as advanced data analytics in the search for a correlation between the DNA region and nucleotide changes in the variants. One may also deploy machine learning algorithms, building a model of the relationship between the variants and other columns in the table. Due to clear and unique labelling, this can be deployed easily.
The last example concerns the data owner. The TCGA-My database provides easy retrieval of the variants and pathways for a patient for future reference or research. The robust architecture of the TCGA-My database also ensures future data can be uploaded smoothly and requires minimal storage and computational resources due to the normalization steps that were deployed during data processing.