Sequence data

We downloaded protein sequences of 67 species with genome sequences from the Joint Genome Institute (JGI) and several other institutions engaged in plant genome sequencing and annotation projects (Supplementary Table S1). For 16 species without genome sequences, we downloaded their expressed sequence tag sequences from UniGene (18) and PlantGDB-assembled unique transcripts from PlantGDB (19), and then built reference proteome for each species (Supplementary Table S2) using a previous established pipeline (17).

Family assignment rules

TFs are usually classified into different families based on their DBDs. We used auxiliary and forbidden domains to distinguish complicated TF families with multiple signature domains. After a comprehensive literature review, we improved the family assignment rules described in the previous version (17) and arranged several families into superfamilies (Figure 2). We removed the forbidden domain Glyco_hydro_14 of the BES1 family, as recent studies demonstrated that BES1 family proteins with this domain also showed TF activity (20).

Open in a separate window

Figure 2.

Refined family assignment rules used for TF identification and assignment. Green ellipses represent TF families and red rectangles represent DBDs. Blue rectangles denote auxiliary domains and purple rectangles denote forbidden domains. Green solid lines link families and DBDs or auxiliary domains and number ‘1’ or ‘2’ indicates number of DBDs. Red dash lines link families and forbidden domains. Families belonging to the same superfamily are arranged within rectangles or rhombi.

Prediction pipeline

We refined the TF prediction pipeline by updating the hidden Markov model (HMM) profiles used to identify TFs and adjusted their thresholds. We downloaded the latest version of HMM profiles from Pfam (version 27.0) (21) for most signature domains and built our own HMM profiles for the remaining domain that did not have available Pfam HMM profiles. We used HMMER 3.0 (22) to identify TFs and assigned them into different families according to the family assignment rules described earlier.

Annotation pipeline

We used a pipeline comprising several packages to annotate identified TFs. Domain structure and GO annotation were predicted by InterProScan (version 4.8) (23). Cross-links to well-known resources were assigned to the best BLAST hits with maximal e-value 1e-10. Nuclear localization signals were predicted by PredictNLS (24). Other information such as expert-curated description, expression, regulation, conserved elements and references was collected from corresponding databases. Multiple sequence alignments (MSAs) for DBDs were constructed by HMM-guided method, and MSAs for full-length protein sequences were inferred by T-coffee (version 9.03) (25). Family trees across 83 species were inferred by FastTree (version 2.1.3) (26) with 100 resamplings. Family trees within each species were inferred by MrBayes (version 3.2.1) (27) based on the Dayhoff model for 50 000 generations. The Help page (http://planttfdb.cbi.pku.edu.cn/help_info.php#tfinfo) describes more detailed information on datasets and parameter settings.

Comprehensive annotations for TFs

A database of well-annotated TFs may provide users with rich information as well as insightful clues for further study. In an attempt to construct a comprehensive knowledgebase for plant TFs, we collected expert-curated description, expression, regulation, mutation and phenotype data from various public resources and made annotations for identified TFs in PlantTFDB 3.0 (Table 3), in addition to abundant annotations provided in the previous two versions (16,17). By integrating information from Entrez Gene (33), UniProtKB (34), GeneRIF (33) and mined by ourselves, we added related references for TFs.

Table 3.

Summary of annotations for TFs in PlantTFDB 3.0

Typea	Species	TF	Entry
Expert-curated description	22	2128	6649
Expression
UniGene	44	44 862	45 239
Microarray	14	15 424	31 975
Plant ontology	5	6850	174 162
Regulation
Binding site/matrix	24	541	729
ChIP-chip/ChIP-seq	1	54	75
microRNA	1	28	43
Hormone	1	417	803
Interaction	10	992	3101
Conserved element	2	3709	63 859
Phenotype	2	4704	147 684
Reference	59	5004	20 255

Open in a separate window

^aNew types of annotations in this version are marked in bold.

Evolutionary conserved elements may work as transcriptional regulatory elements (35,36). Therefore, we collected these elements, which were identified based on the genome alignments of 9 crucifers (36) and 20 angiosperm plants (37), and added them into the current version, in addition to functional genomic annotations described earlier.

Orthologs usually have similar function and are widely used to explore functions of poorly studied proteins. To help users infer the functions of poorly studied TFs, we constructed MSAs and phylogenetic trees within the same family across 83 species, based on conserved DBDs. We further assigned 69 450 TFs into 3924 orthologous groups and constructed phylogenetic trees for each orthologous group. As an aid to decipher their evolutionary relationships, we also built trees for individual TF families within the same species. Hyperlinks to TF pages were added in the tree branches so that the users could browse them conveniently. The MSAs and phylogenetic trees in PlantTFDB 3.0 can be freely downloaded for further analyses. Direct links to TFs of A. thaliana, the best-studied model plant and the best-annotated species in PlantTFDB 3.0, were also generated for all TFs in other species.

TF prediction server

In recent years, the TF classification rules we constructed have been widely used to annotate TFs of newly sequenced genomes (38,39). In this regard, we set up a TF prediction server (http://planttfdb.cbi.pku.edu.cn/prediction.php) for users to identify TFs from their own protein sequences. As A. thaliana is the best-annotated species in PlantTFDB 3.0, links to the best hits in A. thaliana are provided for predicted TFs. Currently, users can upload up to 100 sequences and obtain results within a minute from our server.

The authors thank Joint Genome Institute for the genome annotation of four unpublished species, AGD for Amborella trichopoda, ICGC for Citrus clementina and AGI for three rice species. They also thank their users for their suggestions and comments. They specially thank Ying Dillaha for her language editing of the manuscript.