DBTSS/DBKERO for integrated analysis of transcriptional regulation

Abstract

DBTSS (Database of Transcriptional Start Sites)/DBKERO (Database of Kashiwa Encyclopedia for human genome mutations in Regulatory regions and their Omics contexts) is the database originally initiated with the information of transcriptional start sites and their upstream transcriptional regulatory regions. In recent years, we updated the database to assist users to elucidate biological relevance of the human genome variations or somatic mutations in cancers which may affect the transcriptional regulation. In this update, we facilitate interpretations of disease associated genomic variation, using the Japanese population as a model case. We enriched the genomic variation dataset consisting of the 13,368 individuals collected for various genome-wide association studies and the reference epigenome information in the surrounding regions using a total of 455 epigenome datasets (four tissue types from 67 healthy individuals) collected for the International Human Epigenome Consortium (IHEC). The data directly obtained from the clinical samples was associated with that obtained from various model systems, such as the drug perturbation datasets using cultured cancer cells. Furthermore, we incorporated the results obtained using the newly developed analytical methods, Nanopore/10x Genomics long-read sequencing of the human genome and single cell analyses. The database is made publicly accessible at the URL (http://dbtss.hgc.jp/).

INTRODUCTION

The human genomic variations or mutations in the transcriptional regulatory regions may play roles in the onset and progression of human diseases. In spite of their potential importance and general interests, little is known about how mutations in regulatory regions alter the epigenome or transcription programs, resulting in aberrant cellular phenotypic consequences. In order to understand the effects of these regulatory mutations, the information on the genomic variations needs to be integrated with other omics information, such as epigenome and transcriptome. Since it is sometimes difficult to collect those multi-omics pieces of information directly from clinical samples, information from model experimental systems, such as mouse models and cultured cells, is often used to obtain important clues.

We initiated a database called DBTSS (1), based on our unique full-length cDNA dataset in 2002 (2–4). At the time, we utilized the 5′-end of the full-length cDNA as the precise information of the transcriptional start sites, which can be regulated by alternative promoters. We analyzed the upstream potential promoter regions to outline the genome-wide transcriptional regulations. Since its initial launch, we have made several rounds of updates. In recent years, we associated the transcriptome information with a catalogue of genomic variations, such as public SNP database, as well as the epigenome information, mainly obtained from our own studies, to enable further in-depth analyses on the disease-causing molecular mechanisms. We named this part of the database ‘DBKERO’ (Database of Kashiwa Encyclopedia for human genome mutations in Regulatory regions and their Omics contexts).

In this update, we substantially enhanced both the genomic variation datasets and epigenome variation datasets, particularly focusing on the Japanese population. These datasets were collected as a series of genome-wide association studies and a part of International Human Epigenome Consortium (IHEC) projects (5), respectively. We believe that this should serve as a model case for the genomic, epigenomic and transcriptomic variations occurring in a particular ethnic background that underlie various diseases. Also, we associated the data obtained from clinical samples with that from various model systems, such as the drug perturbation datasets using cultured cancer cells. Furthermore, we incorporated the recent results obtained using the newly developed analytical methods, namely, the long read sequencing analysis using Nanopore (6) and Chromium/GemCode (7), as well as the single cell analyses of cancer cells using the C1 (8,9) and Chromium (10) systems. We believe that these new datasets should be useful for further in-depth analysis of diseased genomes, for which current short read sequencing or bulk sequencing of the materials would give limited information. Statistics of the datasets are shown in Tables 1–3 (for more details, see Supplementary figures or visit our website at http://dbtss.hgc.jp/docs/help_2017.html).

Statistics of the omics datasets

Table 1.

Statistics of the omics datasets

(A) Japanese population
Category	Data source	Number of individuals	References
Germline variation	Human Genome Variation Database (HGVDB) (17 GWAS)	5737 case / 7631 healthy	https://gwas.biosciencedbc.jp/cgi-bin/hvdb/hv_top.cgi
	The Human Genetic Variation Database (HGVD)	1,208	http://www.hgvd.genome.med.kyoto-u.ac.jp/index.html
	Integrative Japanese Genome Variation (iJGVD); ToMMo	1,070	https://ijgvd.megabank.tohoku.ac.jp/
	Japan PGx Data Science Consortium (JPDSC)	2,994	http://www.jpdsc.org/english/
Somatic mutation	Lung adenocarcinoma – National Cancer Center (NCC)	97	PLoS One 2013 8(9) e73484
	Small cell lune cancer - NCC	57	J Thorac Oncol 2014 9(9) 1324-31
	International Cancer Genome Consortium (ICGC) Liver cancer - RIKEN	258	https://dcc.icgc.org/projects/LIRI-JP
	ICGC Liver cancer - NCC	244	https://dcc.icgc.org/projects/LINC-JP
	ICGC Biliary tract cancer	239	https://dcc.icgc.org/projects/BTCA-JP
Normal epigenome	International Human Epigenome Consortium (IHEC) Liver (64 datasets)	8	http://epigenomesportal.ca/ihec/
	IHEC Colon (88 datasets)	11	http://epigenomesportal.ca/ihec/
	IHEC Endometrial (132 datasets)	15	http://epigenomesportal.ca/ihec/
	IHEC Vascular endometrial (4 datasets)	1	http://epigenomesportal.ca/ihec/
(B) World-wide reference datasets
Category	Data source	Number of individuals	References
Germline variation	NCBI dbSNP build 137	***	Nucleic Acids Res 2001 (29) 308-311
	1000 Genomes Project	***	Nature 2015 (526) 68-74; Nature 2015 (526) 75-81
	NHLBI-GO Exome Sequencing Project (ESP)	***	http://evs.gs.washington.edu/EVS/
	Exome Aggregation Consortium (ExAC) (release 0.3)	60,706	Nature 2016 (536) 285-291
Somatic mutation	Catalogue Of Somatic Mutations In Cancer (COSMIC)	***	Nucleic Acids Res 2017 (45) D777-D783
	The Cancer Genome Atlas (TCGA) (11 subtypes)	3,052	Nature Genetics 2013 (45) 1113-1120
	ICGC (43 subtypes)	6,590	Nature 2010 (464) 993-998; http://icgc.org/
Normal epigenome	IHEC (167 datasets)	32	http://epigenomesportal.ca/ihec/
Cancer epigenome	TCGA (2 subtypes)	557	https://cancergenome.nih.gov/
(C) Original datasets of model systems – cell lines and tissues
Category	Number of datasets	Number of samples
Cell line	286	55
Mouse and other organisms	9	5

For the information of the public datasets in this database and references of all of the datasets, also see the statistics page at http://dbtss.hgc.jp/docs/data_contents_2017.html.

Statistics of the drug perturbation datasets

Table 2.

Statistics of the drug perturbation datasets

	Dataset-1	Dataset-2
Sample	5 lung cancer cell lines	23 lung cancer cell lines
Number of compounds	23 + DMSO control	95 + DMSO control
Condition	4 concentration points; 24, 48, 72 h	1 concentration point; 24h
Total number of datasets (RNA-seq)	1299	2011
Total number of datasets (ATAC-seq)	1316	2077

Statistics of the new technologies datasets

Table 3.

Statistics of the new technologies datasets

(A) Single cell analysis
	C1	bead-seq	Chromium (>5k tag)
Sample	4 lung cancer cell lines	5 lung cancer cell lines	5 lung cancer cell lines
Condition	1μM vandetanib; 6h / No treatment	1μM gefitinib; 24h / DMSO control	1μM gefitinib; 24h / DMSO control
Total number of single cells	336	442	47,665
Average number of reads per cell	7,119,082	1,069,847	10,105
(B) Long read analysis
	Chromium/GemCode (whole-exome & regulatory)	Nanopore/MinION (whole-genome)
Sample	23 lung cancer cell lines	4 lung cancer cell lines
Average number of raw reads	45,679,789	451,582
Average depth	53.1 (113.7 Mb)	0.56 (whole-genome)

DATABASE DESCRIPTIONS

New datasets

Genome and epigenome catalogues of the Japanese population

In this update, we integrated several new datasets (Figure 1). The first batch of datasets is the new genome and epigenome datasets collected from clinical samples particularly focusing on the Japanese population. For the genomic variation datasets, we have now included genomic variations of 13,368 Japanese individuals. The raw data was collected from 14 Genome Wide Association Studies (GWAS), which have been conducted around from 2005 to 2015 in Japan (11). A total of 5737 cases and 7631 healthy control samples were enrolled. For the single nucleotide polymorphism (SNP) typing, the SNP arrays from Affymetrix and Illumina (see the web for more details at https://gwas.biosciencedbc.jp/index.html) were used. The frequencies at the corresponding genomic coordinates are represented in the database for each group of the case and control cases. We included genomic variation datasets representing previously published whole-genome/exome sequences (WGS/WES) of a total of 5272 Japanese individuals from three bio-bank cohorts (11–15). These Japanese genomic variations were further associated with external public genomic variation reference datasets. For the germline variations, datasets deposited in the public domains of dbSNP (16) and the Exome Aggregation Consortium (ExAC) (17) databases are used as references. For a reference for the somatic mutations in cancers, the public International Cancer Genome Consortium (ICGC), The Cancer Genome Atlas (TCGA) and the Catalogue of Somatic Mutations in Cancer (COSMIC) databases are used (18–20). Also within the public reference datasets, parts of the data obtained from the Japanese population are highlighted (21,22).

Figure 1.

Overall structure of the database. Overall structure of the database is illustrated. How the Japanese clinical omics information is associated with comprehensive omics information from the model systems is shown. This database also included information newly available from single cell and long read technologies and multi-omics perturbation by chemical compounds. Different categories of datasets are shown in different colors. IHEC: International Human Epigenome Consortium; HGVDB: The human genome variation database; ICGC: International Cancer Genome Consortium; TCGA: The Cancer Genome Atlas; COSMIC: Catalogue Of Somatic Mutations In Cancer; ENCODE: Encyclopedia of DNA Elements; CCLE: Cancer Cell Line Encyclopedia; NHLBI GO-ESP: NHLBI GO Exome Sequencing Project; ExAC: Exome Aggregation Consortium; Roadmap: The NIH Roadmap Epigenomics Mapping Consortium; CMAP: Connectivity Map; GWAS: Genome Wide Association Study.

The genomic variation datasets are associated with the transcriptome and epigenome information datasets (Figure 2A). For the epigenome information, a total of 455 epigenome datasets which were obtained from 67 Japanese individuals were included. These datasets have been collected as the contribution to the IHEC from Japanese groups. Since the Japanese groups were allocated for endometrial, gut, stomach, liver, endothelial, and reproductive organ tissues, in the framework of the IHEC, the data for those organs are specially featured. The Japanese epigenome datasets are associated with various tissues obtained from more than 8000 datasets having different ethnic backgrounds, which have been collected by other IHEC teams worldwide (http://epigenomesportal.ca/ihec/index.html) (23). The Encyclopedia of DNA Elements (ENCODE) (https://www.encodeproject.org/) (24) and the NIH Roadmap Epigenomics Mapping Consortium (Roadmap) (25) datasets were also associated as the reference dataset. Those collective epigenome datasets are further associated with a total of 286 genome, epigenome and transcriptome datasets, which have been published in previous rounds of the updates of DBTSS/DBKERO (Table 1). To our knowledge, this update has made our database the richest resource of the multi-omics information focusing on the Japanese population.

Figure 2.

The overview of the genome viewer. (A) A representative view of the genome browser harboring standard omics information from Japanese clinical samples. Data for the indicated layers of the omics analyses is shown. The information around the BRAF gene is represented for a Japanese case included in the IHEC dataset. Gene expression, histone modification and DNA methylation patterns are displayed in the indicated tracks. Variant frequencies are also shown for the Japanese SNPs from GWAS datasets. (B) An example of the SPARQL search for connecting the search to the UniProt database for the BRAF gene. For more details, see Supplementary Figure S1. (C) Drug perturbation viewer. The viewer represents the distribution of fold expression in mRNA and chromatin accessibility changes in the regulatory regions in response to the drug treatments. (D, E) Viewers for the new analytical methods. (D) The single cell viewer represents the gene expression diversities in each cell, which were obtained from the C1 and Chromium platform. The user can switch the interface to that of cellular population diversities on the two-dimensional plot, which were obtained from the Chromium platform. (E) The long read viewer represents phasing information in cancer cell lines. The phasing information obtained from the Nanopore whole-genome sequencing and the Chromium/GemCode linked reads is shown as indicated.

RDF as a further integration clue for other databases

To further expedite the data integration with outer databases, we employed RDF (Resource Description Framework) as the data model. Even not for all, but a significant part of the data can be described in the RDF format. Therefore, the database search from the above datasets are possible from the outside based on the SPARQL (Simple Protocol and RDF Query Language) search. In Figure 2B, the SPARQL search between the DBTSS/DBKERO and the UniProt database (26) at the Ensembl database (27) is exemplified (also see Supplementary Figure S1). This interface, which enables the data search across the databases world-wide, should be useful to complement the data lacking from this database, such as proteome and metabolome.

Omics datasets from experimental models

One of the unique features of our database lies in the fact that we focus on the experimental model systems. We believe that model systems would give an important clue about the biological interpretation of the clinical omics information. The use of model systems is particularly important when clinical samples cannot be directly used for hypothesis-generation or for validation analyses. Genetic or drug perturbations are frequently impossible using clinical samples. In addition to datasets of below human cell lines, we generated a series of multi-omics datasets of the mouse model as a model of particular human diseases. The mouse datasets were connected to the human datasets via our genome-genome alignment (28). The human-mouse inter-species association of the tissues depending on the disease models is also represented in a search engine.

Among a series of datasets from the model systems, we would like to especially draw attention to our newly published datasets on the drug perturbations of the cancer cell lines (Figure 2C). For lung cancer cell lines (half are from Japanese origins and half from other ethnic backgrounds), a comprehensive multi-omics catalogue has been generated and represented (WGS; DNA methylation by bisulfite sequencing; epigenome information represented by eight types of histone modifications; open chromatin by ATAC-seq; transcriptome sequencing of mRNA and miRNA and their transcriptional start sites; long-read sequencing; and single cell transcriptome) (8,29). For these cell lines, 95 representative drugs were administered at varying concentrations and time points (Supplementary Figure S2). RNA-seq and ATAC-seq were conducted to monitor transcriptome and epigenome responses of the cells, resulting in a total of 3240 RNA-seq and 3393 ATAC-seq datasets. Summary of the data contents for this part of the database is shown in Table 2. We associated these original profiles with the datasets from the Connectivity MAP (CMAP), which are publicly available datasets for the similar purpose (30). These datasets should be helpful when the user wants to know possible consequences of the drugs administration in a given mutation types or epigenome/transcriptome types in in vivo cases.

Dataset obtained from new analytical platforms

In this update, we also attempted to further enrich the multi-omics information using the latest genomic technologies (Figure 2D and E). Recently, several new analytical platforms have been developed and the data production has been started in our laboratories as well as world-wide. We particularly focused on the single cell sequencing technologies and the long-read sequencing technologies. These new methods would shed new light on the previously uncharacterized molecular mechanisms of the diseased genomes. However, these technologies are still incomplete and widely-used current platforms usually have distinct advantages and disadvantages. We believe that it would be extremely useful if the data obtained from the same materials using the different platforms are integrated and simultaneously represented as a database.

Single cell analysis

Recent single-cell sequencing technologies have opened the possibility of analyzing individual cells. A number of reports have demonstrated that single-cell analysis provides pivotal information for elucidating cellular plasticity and diversity within a given cellular population in vitro and in vivo. Cancer consists of various cell types, including cancer cells, cancer-associated fibroblasts, tumor-infiltrating leukocytes and vascular cells. When we consider the drug perturbations of cancer, we have to inevitably consider the drug response of each population of the cells and their mutual dependence. Analyses at the single-cell resolution is indispensable in case of disease and targets for the treatments occur in a limited number of cells or when the cellular micro-environments play an essential role.

From a technical perspective, the currently used single cell analytical methods can be separated into two categories. In the first method, cells are physically separated by microfluidics, FACS, or other methods, followed by the reactions of reverse transcriptions and PCR amplifications in an individual reaction chamber manually or with the aid of robotics (9). In the second method, cells are confined to a micro-droplet, and individual cells are separately marked using molecular barcoding technology (10). The representative platforms of these methods are the C1 system of Fluidigm and bead-seq system (31), and the Chromium system of 10x Genomic, respectively. Although both platforms are commonly employed for single cell analyses for a wide variety of cells, each of the platforms has several intrinsic advantages and disadvantages. Namely, the C1 system yields sufficient sequence coverage per cell but sparse coverage for the cellular population. The converse is the case for the Chromium system.

In this database, single cell RNA-seq datasets obtained from a series of cancer cells are represented. The data was collected using the C1, bead-seq and Chromium platforms (Table 3A). For the materials, we used the same cancer cell materials as described above. We found it extremely informative to integrate a wide variety of datasets to provide a full spectrum of information for the single-cell. Particularly, as exemplified in Figure 2D, the C1 datasets were used to precisely represent the information of the single cells while the Chromium datasets were used to estimate the frequency of the cells having the corresponding expression patterns within the population. Also see the ‘Example Tour’ below for further concrete ideas.

Long read sequencing

Other datasets obtained from ‘the latest analytical methods’ are the long read sequencing dataset (Table 3B). In the example of cancers, the functional relevance of the mutations occurring in the regulatory regions still remains mostly elusive. One of the largest drawbacks preventing its more efficient characterization lies in the fact that the distance between the transcriptional regulatory region and the regulated gene region is occasionally beyond the reach of the short-read sequencing. Therefore, it is generally difficult to identify the allelic context of ‘regulatory’ mutations, which is how the mutation, usually occurring in one of the loci, affects the target gene on the same chromosome. As exemplified in Figure 2E, by utilizing the long read sequencing technologies, the single nucleotide variants (SNVs) in the regulatory regions can be phased to the downstream heterozygous SNPs/SNVs in the coding regions, if they are present. Once phased, we found it possible to investigate their transcriptomic consequences by examining whether the ChIP-seq variant tags of the regulatory SNVs and the RNA-seq variant tags of their target transcripts showed biased frequency between the mutant and reference alleles (also see the below ‘Example Tour’).

In this update, we included the long-read sequencing datasets obtained from the Chromium/GemCode system of 10× Genomics and the MinION sequencer of the Oxford Nanopore Technologies. While the Chromium/GemCode method enables long-read sequencing by intensive use of the barcoding technologies and bioinformatics (so-called ‘synthetic long-read (linked read)’ method), the MinION enables it by the ‘physical long-read’ method. The former method has advantages in the cost, but has a disadvantage in leaving some ambiguity in the resulting phasing information. Particularly, the erroneous phasing is inevitable for aneuploid regions or copy number aberration regions in cancer genomes. On the other hand, the MinION method does not have such a disadvantage, instead, its low base-call quality and high sequencing cost impose considerable barrier preventing its sole use for the WGS (note: we employed the MinION to represent the physical long read method, but PacBio sequencing is an alternative). Within this particular dataset, by mutually complementing the drawbacks in the respective methods, we identified 137 potential regulatory mutations affecting transcriptional regulation. Among them, 84 SNVs could create and/or disrupt potential transcription factor binding sites.

Unique features in this dataset

Another important unique feature of this database is that intensive data collection was made by a wide variety of analytical methods focusing on a group of model systems, such as lung cancer cells. We also designed the databases by arranging such information from the model systems to collectively enable interpretation of the information obtained from clinical samples, for which complete suites of omics datasets are mostly unavailable. Particularly, for a series of cancer cells, the datasets, covering whole genome, epigenome, transcriptome data in bulk cells as well as long-read sequencing and single cell data in response to various experimental conditions, have been collected using exactly the same materials. To our knowledge, this is the richest datasets for which most, if not all, of the currently available omics analyses have been intensively conducted. We hope these datasets will serve as a valuable resource to obtain a comprehensive omics view in a given cellular entity from a systems biology viewpoint, namely, how the genomics mutations affect the transcriptional programs or how such transcriptomic perturbation observed in bulk could be dissected down to single cell components. Such cellular information should be further interpreted in the context of the clinical samples having similar genomic, epigenomic or transcriptomic aberrations at least in a sub-module in the genomic system.

AN EXAMPLE TOUR OF DATA RETRIEVAL

Figure 3 illustrates the model tour for the usage of the database. A user may start the search of the database with a somatic mutation found in the promoter region of the TERT gene in Japanese liver cancer patients (chr5:1295113, G>A) (https://dcc.icgc.org/projects/LINC-JP) (Figure 3A; see Supplementary Figure S3 for the more detailed guide). This gene is the key regulator of telomere stability. The search would find that this is a highly frequent mutation in cancers but never represented in germline variations. Inspection of the transcriptome and epigenome information in the surrounding region suggests that this mutation is located in the transcriptional regulatory region, which regulates the transcription of the TERT gene. When further information is sought in the datasets of the model cancer cell lines, the user would find that there are some cell lines harboring the mutation at exactly the same genomic locus and the transcription of the gene body is aberrantly enhanced, thus, could be used as a model to analyze its biological functions. Indeed, TERT promoter mutations have been detected in melanoma (32,33), liver cancer (34) and other various types of cancers according to the COSMIC database.

Figure 3.

Example tours of the TERT and MT1A genes. (A) The tours of the TERT gene. The guide how to show the similar results in the web interface is illustrated. Further detailed guide of the same tour is also shown at the help page (http://dbtss.hgc.jp/docs/help_2017.html). For the tour, follow the link as illustrated in Supplementary Figure S3: 1. Input ‘TERT’ to the keyword field box at the top left part of the top page. 2. In the genome browser, use the track buttons to display the clinical dataset of ‘ICGC Liver Cancer-NCC, Japan’ by selecting ‘Japanese,’ ‘Genome,’ ‘Cancer cell,’ ‘Clinical samples’ and ‘IGCG’ track in the ‘Standard multi-omics data.’ Also add the datasets of germline variation by selecting ‘Japanese,’ ‘Genome,’ ‘Normal cell’ and ‘Clinical samples’ tracks in the menu of ‘Standard multi-omics data.’ Add the germline variations in other ethnic groups by selecting the ‘worldwide’ track. For mutations in cancer cell lines, select SNVs of the indicated cell lines. Find a mutation (chr5:1,295,113, G>A) in the TERT promoter region. View the results in the upper right panel of the figure. 3. Display multi-layered data of the cell lines. Select SNVs, H3K4me3 and Pol II ChIP-seq patterns, TSS, and rpkm values of RNA-seq for the indicated cell lines. Note for Figure 3A: to directly visit each of the panels, follow the links as below: ♦ http://dbtss.hgc.jp/#kero:chr5:1295007-1295133&initShow=sequence,cpg,refGene,snp_icgc_LINCJP,snp_pgx,osnp10,snp_dbsnp137,snp_ESP137,snp_1000genome,snv_RERFLCad1,snv_RERFLCOK; ♦ http://dbtss.hgc.jp/#kero:chr5:1292755-1295691&initShow=sequence,cpg,refGene,snv_RERFLCad1,snv_RERFLCOK,snv_PC9,peak_RERFLCad1_H3K4me3,peak_RERFLCOK_H3K4me3,peak_PC9_H3K4me3,peak_RERFLCad1_PolII,peak_RERFLCOK_PolII,peak_PC9_PolII,tss_RERFLCad1,tss_RERFLCOK,tss_PC9,rpkm_RERFLCad1,rpkm_RERFLCOK,rpkm_PC9 (B) The tours of the MT1A gene. The guide how to show the similar results in the database are illustrated. For more details see the web (http://dbtss.hgc.jp/docs/help_2017.html). Follow the link as illustrated in Supplementary Figure S4: 1. Input ‘MT1A’ to the keyword field box at the top left part of the top page. 2. Select the ‘GemCode Phasing Patterns’ of the H1975 cell line. Find a mutation (chr16:56,638,440, C>CG) in the haplotype 2 of the MT1A upstream region (upper left panel). 3. Add epigenome and transcriptome information of the H1975 cell line around the mutation (upper right panel on blue background). Select H3K27ac ChIP-seq and DNA methylation of BS-seq for epigenome patterns, and TSS and rpkm of RNA-seq for transcriptome patterns. 4. To view the data of expression variation in individual single cells, display the rpkm/ppm distribution of the C1, bead-seq and Chromium single cell platforms. To view the distribution of the expression levels of the MT1A gene in each cell, select the C1 system (lower left panel on yellow background). For information of a large number of cells, select the Chromium system. Go to the single cell viewer from the summary link and see the expression variation of MT1A gene on the two dimensional t-SNE plot (lower right panel on yellow background). Note for Figure 3B: to directly visit each of the panels, follow the links as below; ♦ http://dbtss.hgc.jp/#kero:chr16:56638666-56640088&initShow=sequence,cpg,refGene,gemcode_H1975 ♦ http://dbtss.hgc.jp/#kero:chr16:56638427-56638490&initShow=sequence,cpg,refGene,gemcode_H1975 ♦ http://dbtss.hgc.jp/#kero:chr16:56,638,427-56,638,490&initShow=sequence,cpg,refGene,gemcode_H1975,peak_H1975_H3K27ac,bs_H1975,tss_H1975,rpkm_H1975 ♦ http://dbtss.hgc.jp/#kero:chr16:56,638,077-56,639,041&initShow=sequence,cpg,refGene,gemcode_H1975,peak_H1975_H3K27ac,bs_H1975,tss_H1975,rpkm_H1975 ♦ http://dbtss.hgc.jp/#kero:chr16:56,638,381-56,640,372&initShow=sequence,cpg,refGene,ppmdist_c1_PC9,ppmdist_c1_LC2ad_2,ppmdist_pc9_dmso,ppmdist_pc9_gefitinib.

In another example, a user may start with a mutation found in the upstream region of the MT1A gene (chr16:56638440, C>CG) in H1975 lung cancer cell line (Figure 3B; see Supplementary Figure S4 for the more detailed guide). This gene is a member of the metallothioneins. It is reported that the mRNA/protein levels of this gene are associated with cellular proliferation and migration, differentiation and chemoresistance of cancers (35) as well as for genes from related families (36,37). The user would further connect the searches to find that this enhancer mutation is phased to its transcript region. The ChIP-seq tags for the active histone marks and the transcript RNA-seq tags show a biased representation. When the user examines the single-cell datasets, he/she will also find that the gene expression is diverse among individual cells, suggesting the transcriptional heterogeneity and diverse response to stimulation among cancer cells, which may give essential information, when drug intervention on this gene is contemplated.

AVAILABILITY

A detailed user manual is available on the website (http://dbtss.hgc.jp/docs/help_2017.html). Documents on data processing and information of quality control are also provided at the page of the experimental procedures (http://dbtss.hgc.jp/?doc:protocol_2017.html). Statistics for the current database are also presented in the statistics section (for more details, visit the statistics page at http://dbtss.hgc.jp/docs/data_contents_2017.html). All of the short read sequences used for the database are those which have been deposited in the Short Read Archives and JGA Database for Control Access in DDBJ (http://www.ddbj.nig.ac.jp/index-e.html). Accession numbers are as appear in the statistics section (left frame in the top page).

CONCLUSIONS

In this database, we attempt to develop a versatile database platform. We expect this database to facilitate the analyses regarding how germline variations or somatic mutations in cancers residing in transcriptional regulatory regions may affect the transcriptional regulation of their target genes in the diseased genome contexts. Since it is still generally difficult to collect the complete omics datasets directly from clinical samples, the surrogate data from model systems, such as cultured cells and mouse models, play important roles. In this database, we arranged the retrieval systems so that information from clinical samples should be easily associated with wider variation of omics information in model systems.

For the eventual diseased genomes, we particularly focused on those from the Japanese population. However, the framework of the database should be similarly useful by substituting the genomic information by other datasets of a different ethnic background. Ethnic backgrounds may potentially have substantial effects on the disease etiology. By providing the solid omics data core and referring them to the clinical samples, we believe our database can help a broad range of users who study clinical relevance of the genomic variations/mutations in genes and particularly in the regulatory regions.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR online.

ACKNOWLEDGEMENTS

The database is maintained on the supercomputer system at Human Genome Center, the Institute of Medical Science, the University of Tokyo. We thank the members of the CREST/IHEC and other IHEC teams worldwide. We are also grateful to H. Wakaguri, Y. Kuze and T. Horiuchi at the University of Tokyo and the technical staffs of Database Center for Life Science.

FUNDING

DBTSS/DBKERO is financially supported with a Grant-in-Aid for Publication of Scientific Research Results (Databases) by Japan Society for the Promotion of Science, a Grant-in-aid for Scientific Research on Innovative Areas ‘Platform for Advanced Genome Science’ [16H06279] from the Ministry of Education, Culture, Sports, Science and Technology of Japan; Database Integration Coordination Program from Japan Science and Technology Agency; CREST/IHEC, Japan Agency for Medical Research and Development. Funding for open access charge: CREST/IHEC, Japan Agency for Medical Research and Development.

Conflict of interest statement. None declared.

Author notes

These authors contributed equally to this work as first authors.