H-InvDB in 2013: an omics study platform for human functional gene and transcript discovery.

H-InvDB (http://www.h-invitational.jp/) is a comprehensive human gene database started in 2004. In the latest version, H-InvDB 8.0, a total of 244 709 human complementary DNA was mapped onto the hg19 reference genome and 43 829 gene loci, including nonprotein-coding ones, were identified. Of these loci, 35 631 were identified as potential protein-coding genes, and 22 898 of these were identical to known genes. In our analysis, 19 309 annotated genes were specific to H-InvDB and not found in RefSeq and Ensembl. In fact, 233 genes of the 19 309 turned out to have protein functions in this version of H-InvDB; they were annotated as unknown protein functions in the previous version. Furthermore, 11 genes were identified as known Mendelian disorder genes. It is advantageous that many biologically functional genes are hidden in the H-InvDB unique genes. As large-scale proteomic projects have been conducted to elucidate the functions of all human proteins, we have enhanced the proteomic information with an advanced protein view and new subdatabase of protein complexes (Protein Complex Database with quality index). We propose that H-InvDB is an important resource for finding novel candidate targets for medical care and drug development.

INTRODUCTION

Along with the sequencing of the first human reference genome (1), several lines of human transcriptome study using a large number of validated human transcripts were carried out. As full-length complementary DNA (cDNA) is the ideal resource for the study, our consortium aimed to collect human full-length cDNA sequenced by four projects: Full-Length cDNA Japan (FLJ) (2), Human Unidentified Gene-Encoded Large Proteins (HUGE) (3), Mammalian Gene Collection (MGC) (4) and Munich Information Centre for Protein Sequences (MIPS) (5). These projects were conducted at five institutions: New Energy and Industrial Technology Development Organization (NEDO), Kazusa DNA Research Institute (KDRI), the National Institutes of Health (NIH, USA), German Research Centre for Environment and Health (GSF) and Chinese National Human Genome Centre (CHGC) (6). Our consortium then held an international workshop called Human Full-Length cDNA Annotation Invitational (H-Invitational or H-Inv) to manually annotate the registered human full-length cDNA sequences on our annotation system by expert scientists and annotators (7). To release the annotation results, the first H-InvDB was constructed in 2004, and as of the third version in 2006, H-InvDB was extended to include all published human cDNA in addition to H-Inv human full-length cDNA (8).

At present, H-InvDB has been developed as not only a human transcriptome database but also one of the largest integrative human omics databases available to human gene researchers in various biological fields. One of the features of H-InvDB is that all published human cDNA sequences were annotated by a rigorous annotation pipeline confirmed at H-Invitational (7,9). For example, we examine sequence quality, sequence identity with the human reference genome sequence, sequence orientation (some cDNA sequences are registered in reverse direction), chimeric or truncated cDNAs and possible contamination from other species. Thus, most artifacts were removed and misannotations were expected to be few. H-InvDB also contains several specific H-Inv sub and satellite databases based on the annotation of H-Inv human transcripts (Figure 1). Databases involving gene expression (H-ANGEL) (10), molecular evolution (Evola) (11), genetic polymorphism (VarySysDB) (12) and alternative splicing (H-DBAS) (13) have been developed. Thus, users can find objective human annotation information in diverse combinations by using the search system of H-InvDB. In addition to these databases, H-InvDB is also connected with external databases by the web service application program interfaces (APIs) and Hyperlink Management System (HMS) (14). On these accounts, H-InvDB is a reliable and useful database for omics studies.

Figure 1.

A schematic diagram of H-InvDB as a central hub for human omics study. Each content is described shortly in the Quick guide page (http://h-invitational.jp/hinv/ahg-db/tools.jsp).

CHARACTERISTICS OF H-InvDB RELEASE 8.0

Update information

In the latest version of H-InvDB 8.0, 244 709 human transcript sequences extracted from DDBJ (15) were freshly mapped on the assembled reference genome UCSC hg19 (16). Clustering the transcripts revealed 43 829 gene loci called H-Inv clusters (HIXs) (Table 1). Among these 43 829 genes, 35 631 were predicted as potential protein-coding genes. This number is much larger than the number of nonredundant protein entries in UniProtKB/SwissProt (17), which is a literature-based, human curated database of known proteins, because H-InvDB contains both known and predicted proteins from human transcripts. We classified them into seven protein categories according to the strength of protein evidence (7) and found that 22 898 genes were predicted to have at least one protein functional motif (Categories I–III) (Table 2).

Table 1.

Statistics of H-InvDB 8.0

Number of gene clusters (HIX)	Number of transcripts (HIT)	Number of proteins (HIP)
43 829	244 709	147 684

Table 2.

Statistics of representative HIPs

Category	Definition	Number of representative HITs
I	Identical to known human protein (≥98% identity and 100% coverage)	16 128
II	Similar to known protein (≥50% identity and ≥50% coverage)	5872
III	InterPro domain containing protein	898
IV	Conserved hypothetical protein	1705
V	Hypothetical protein	5268
VI	Hypothetical short protein (20–79 amino acids)	5068
VII	Pseudogene candidates	692
Total		35 631

Including all these protein categories, all H-Inv transcripts (HITs) were annotated with various sequence features, such as gene structures, alternative splicing variants, noncoding functional RNA, protein functions, functional domains, subcellular localizations, metabolic pathways, protein 3D structure, genetic polymorphisms (single-nucleotide polymorphism, indels and microsatellite repeats), association with diseases, gene expression profiling, molecular evolutionary features, protein–protein interactions (PPIs) and gene families/groups. These annotations were assigned to not only H-InvDB but also the corresponding specific H-Inv sub and satellite databases in detail. These annotations are also used as search items in the H-InvDB Navi system (8) for compound retrieval. Among the H-Inv satellite databases, H-InvDB Enrichment Analysis Tool (HEAT) (8) was considerably upgraded. HEAT is a tool for gene-set enrichment analysis based on various annotation in H-InvDB, such as InterPro (18), GO (19), KEGG pathway (20), SCOP (21), subcellular localization, chromosomal band, gene family and tissue specific expression in H-ANGEL (10). It searches for H-InvDB annotations that are significantly enriched in a user-defined gene sets as compared with the entire H-InvDB representative protein-coding transcripts. We newly added promoter motifs of all human genes based on JASPAR (22) and PPIs in the HEAT system. This enabled us to conduct extensive data mining with the HEAT system.

Advantages of H-InvDB

We compared 43 829 H-Inv genes with RefSeq (23) and Ensembl (24) genes to enumerate the numbers of unique and overlapping entries. Although the numbers of H-Inv unique genes were similar to those of the Ensembl unique ones (19 309 and 19 063, respectively) (Figure 2A), H-InvDB uses only rigorously annotated human cDNA sequences, including those of experimentally validated full-length cDNA (7). These characteristics suggest that our uniquely annotated genes were likely to be biologically functional. To investigate the evidence for protein coding of H-Inv unique genes, we also compared frequencies of the genes in protein categories between H-InvDB and the consensus coding sequence (CCDS) (25) (Figure 2B). The result indicated that unknown functional proteins (Categories V and VI) and nonprotein-coding sequences were frequent in H-Inv unique genes. As described earlier, these H-Inv unique unknown proteins were completely transcribed as they can indeed have some functions. In fact, 233 genes, which have been classified as hypothetical proteins (Categories V and VI) in the previous version of H-InvDB 6.2, turned out to be functional proteins (Categories I–III) in the latest version of H-InvDB 8.0, because they were found in Online Mendelian Inheritance in Man (OMIM) (26) (Table 3). Among them, 11 Category I genes were suggested to associate with Mendelian disorders based on OMIM (Supplementary Table S1). Two of the 11 genes were annotated as Waldenstrom’s macroglobulinemia susceptibility and other two were annotated as psoriasis susceptibility. In addition, 11 genes that have been classified as hypothetical proteins (Categories V and VI) in H-InvDB 6.2 turned out to be noncoding RNA candidates (Supplementary Table S2). Four of these genes were annotated as similar to functional noncoding RNAs.

Figure 2.

Comparison of gene numbers between H-InvDB and other databases. (A) The Venn diagram represents the numbers of unique and overlapping genes among H-InvDB, RefSeq and Ensembl. (B) The bar graph represents the numbers of H-Inv unique genes when compared with CCDS genes. The roman numerals indicate protein categories shown in Table 2.

Table 3.

Protein category-upgraded genes relating with Mendelian disorders in only H-InvDB 8.0

Categorya	Number of category-upgraded genes
Upgrade from V or VI to I	11
Upgrade from V or VI to II	209
Upgrade from V or VI to III	13

aDefinition of category is shown in Table 2.

New features

We had originally developed annotation viewers for transcriptomes and genomes, called ‘Transcript view’ and ‘Locus view’, respectively. In addition to these viewers, we provided a new viewer named ‘Protein view’ for the annotations of the human proteome (Figure 3A). In Protein view, annotation information of H-Inv proteins (HIPs) is provided. Furthermore, through the web service APIs, a link to GlycoProtein DataBase (27) is added and the glycosylation site is illustrated in the figure of Protein view. As human cDNA clones are necessary for protein expression experiments, we added links to the human cDNA clone databases such as Biological Resource Center (NBRC) and Human Gene and Protein Database (HGPD) (28), which are connected by HMS. Using these links, users can access the databases and obtain actual human cDNA clones for various experiments. A new subdatabase was also constructed and connected with H-InvDB. This new subdatabase called Protein Complex Database with quality index (PCDq) (29) is a human protein complex database with complex quality index, which describes evidence levels as subunits (protein members) of the protein complex. From the human PPI network dataset integrated from the six PPI data, human protein complexes were predicted and curated with the literature. Thus, PCDq consists of both known and predicted complexes/subunits (Figure 3B). PCDq is expected to enable users to investigate protein interactions in more detail by protein subunit rather than whole protein.

Figure 3.

Screenshot of a part of protein view and the top page of PCDq. (A) Hyperlinks to NBRC and HGPD are shown in a red circle. (B) Entrance to PCDq is http://www.h-invitational.jp/hinv/pcdq/.

FUTURE PERSPECTIVES

At present, the identification of all human proteins is proceeding worldwide. H-InvDB will continue to offer tools for proteome studies. For example, we are now collecting information on posttranslational modification. Using feedback from various experimental results at the protein level, we intend to develop H-InvDB as the best central hub for human omics study. In addition, personal genome annotation such as the prediction of disease susceptibility using individual gene mutations will be much required. Therefore, we intend to expand the field of personal genomics in future. In addition to the web service APIs of the present H-InvDB, we will provide annotation data in the Resource Description Framework (RDF) (http://www.w3.org/RDF/). We aim to improve the efficiency of accessing molecular biological data by integrating international databases in a more sophisticated manner using this semantic web technology.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online: Supplementary Tables 1 and 2.

FUNDING

National Institute of Advanced Industrial Science and Technology (AIST) and the life science database project of the Ministry of Economy, Trade, and Industry (METI) of Japan. Funding for open access charge: AIST.

Conflict of interest statement. None declared.