TECHNOLOGY AND CODE article

Front. Genet., 11 September 2020

Sec. Computational Genomics

Volume 11 - 2020 | https://doi.org/10.3389/fgene.2020.566569

EAT-UpTF: Enrichment Analysis Tool for Upstream Transcription Factors of a Group of Plant Genes

  • 1. Department of Chemistry, Seoul National University, Seoul, South Korea

  • 2. Plant Genomics and Breeding Institute, Seoul National University, Seoul, South Korea

  • 3. Research Institute of Basic Sciences, Seoul National University, Seoul, South Korea

Abstract

EAT-UpTF (Enrichment Analysis Tool for Upstream Transcription Factors of a group of plant genes) is an open-source Python script that analyzes the enrichment of upstream transcription factors (TFs) in a group of genes-of-interest (GOIs). EAT-UpTF utilizes genome-wide lists of TF-target genes generated by DNA affinity purification followed by sequencing (DAP-seq) or chromatin immunoprecipitation followed by sequencing (ChIP-seq). Unlike previous methods based on the two-step prediction of cis-motifs and DNA-element-binding TFs, our EAT-UpTF analysis enabled a one-step identification of enriched upstream TFs in a set of GOIs using lists of empirically determined TF-target genes. The tool is designed particularly for plant researches, due to the lack of analytic tools for upstream TF enrichment, and available at https://github.com/sangreashim/EAT-UpTF and http://chromatindynamics.snu.ac.kr:8080/EatupTF.

Introduction

The rapid development of high-throughput technologies such as RNA sequencing (RNA-seq), DNA affinity purification followed by sequencing (DAP-seq), and chromatin immunoprecipitation followed by sequencing (ChIP-seq) has led to an explosion in the availability of sequence data. The high-throughput analyses produce lists of genes that are under a particular regulation. When such lists are generated, researchers usually try to understand the biological implications of groups of genes-of-interest (GOIs). To this end, routine follow-up studies typically include gene ontology (GO) enrichment analyses (; ) and Kyoto Encyclopedia of Genes and Genomes (KEGG) mapping (). In addition, transcription factor (TF) prediction analyses (; ) can be performed to identify consensus upstream regulators of a subset of GOIs, giving a biological insight into the integrated role of the genes under specific conditions. Furthermore, comprehensive identification of TF binding sites and cognate TFs can be used to characterize regulatory networks containing GOIs.

Several bioinformatics tools have been developed to predict upstream TFs. The cis-element sequences that are commonly conserved in sets of input query genes can be identified using ab initio motif enrichment algorithms such as MEME (). The identified consensus sequences can be further analyzed to compare enrichment of TF candidates to the consensus binding motifs provided by databases of experimentally validated TF binding sites, such as JASPAR () and TRANSFAC (). Recently, accumulating data have enabled that position weight matrix (PWM)-based enrichment methods solely cover a wide range of upstream TF prediction. This theoretical basis has been implemented in various upstream TF prediction tools, such as TFEA.ChIP, oPOSSUM, and PlantRegMap (; ; ). However, this approach occasionally produces a considerable number of false positives due to short and degenerate nature of TF-binding sites (). In addition, this method is complicated by the fact that TFs can sometimes bind to gene sequences that differ from their consensus binding sites, and that several TFs undergo protein–protein interactions that enable them to recognize additional DNA sequence motifs. Overall, it is clear that a simplified and realistic prediction of TFs controlling a group of GOIs is necessary to generate a confident conclusion.

In this regard, several bioinformatics tools implementing TF enrichment analysis have been developed using ChIP-seq datasets (; ; ). However, these tools are applicable mainly to animal systems, and no codes have been released to analyze enriched upstream TFs for other species. Based on explosive accumulation of plant DAP-seq and ChIP-seq data, there are growing needs to integrate the NGS data and use them to retrieve upstream TFs in plant researches. Notably, O’Malley and colleagues adapted the innovative DAP-seq method and have successfully produced a genome-wide collection of target genes for 349 TFs in Arabidopsis thaliana (). In this study, we have developed the “Enrichment Analysis Tool for Upstream Transcription Factors of a group of plant genes” (EAT-UpTF) tool to provide upstream TF enrichment analysis (). As a proof of concept, we combined it with the Arabidopsis DAP-seq database to analyze the enrichment of upstream TFs in a group of Arabidopsis GOIs. We found that EAT-UpTF was able to robustly evaluate the over-representation of experimentally validated upstream TFs binding to a group of GOIs without the prediction of cis-motifs.

Methods

High-throughput sequencing analyses typically produce sets of GOIs that require further analyses to evaluate their biological implication and underlying regulatory mechanisms. EAT-UpTF is linked to a DAP-seq database (Plant Cistrome database1) that provides a list of TF-target genes (locus IDs). When a set of GOIs is input in the form of locus IDs, EAT-UpTF identifies the TF targets and compares their relative enrichment in the list of GOIs with that in the total genomic genes. As a result, target genes of certain TFs, which are enriched (over-represented) in the set of GOIs can be identified as a major upstream regulators of the gene group (Figure 1). To examine the statistical significance of over-representation, the SciPy module () is used to perform hypergeometric and binomial tests, which differ in that the binomial test considers replacement whereas the hypergeometric test does not. These two tests are used to compare the occurrence of x genes (a subset of TF-target genes) among n genes (GOIs) with that of X genes (total TF-target genes) among N genes (total reference genes). Comparisons with relatively large differences (x/n – X/N) can then be considered to identify upstream TFs that may play a particular role in regulating at least a subset of GOIs.

FIGURE 1

For the initial validation of EAT-UpTF, we used the DAP-seq Arabidopsis database, which lists the target genes of a vast majority of Arabidopsis TFs (∼349). Since EAT-UpTF performs enrichment analyses for hundreds of TFs simultaneously, a post hoc test should be applied to counteract the type I errors (false positives) originating from multiple testing. A number of post hoc analyses can be used to compensate for the increase in the false positive rate caused by multiple tests. The most widely used method is the family-wise error rate (FWER) correction, named after Carlo Emilio Bonferroni. The Bonferroni correction tests individual hypotheses at a significance level of a/m, where a is the desirable alpha level and m is the number of tests performed (; ). This correction method is considered conservative when a large number of tests are conducted, but was likely appropriate in our analysis because the multiple hypothesis tests were limited to several hundreds of TFs. Another post hoc analysis option is the false discovery rate (FDR) correction described by . The Benjamini-Hochberg FDR correction tests hypotheses at a significance level of ka/m, where a is the desirable alpha level, m is the number of tests performed, and k is the rank of the p-value of the hypothesis. These two most popular post hoc analyses have been implemented in the current version of EAT-UpTF using the Statsmodels module of Python ().

Results and Discussion

To validate the relevance of EAT-UpTF, we input a gene set bound by the LATE ELONGATED HYPOCOTYL (LHY) TF in Arabidopsis, which was identified via a ChIP-seq analysis (). EAT-UpTF identified LHY as being an over-represented upstream TF in the test set. Specifically, 71.6% of the input genes were retrieved to be bound by LHY (Table 1) and LHY was identified as one of the top five enriched TFs in the test set (Table 1). The mismatch between the EAT-UpTF output and the ChIP-seq data might be related to the fact that DAP-seq is generally more stringent than ChIP-seq. Typically, DAP-seq produces a rigorous gene set and usually identifies a smaller number of TF-target genes than ChIP-seq. Indeed, all of the LHY-target genes identified by DAP-seq were included in the list of LHY-target genes identified by ChIP-seq analysis.

TABLE 1

TF ID (AGI ID)xanbObserved (%)XcNdExpected (%)p-ValueCorrected p-valueeGene symbolsGene names
AT5G0284028772239.84,11027,20615.15.84 × 10–602.04 × 10–57LCL1LHY/CCA1-LIKE 1
AT3G0960042672259.08,27627,20630.42.59 × 10–584.52 × 10–56RVE8, LCL5LHY-CCA1-LIKE5, REVEILLE 8
AT3G5685027572238.13,93627,20614.56.43 × 10–577.48 × 10–55AREB3, DPBF3ABA-RESPONSIVE ELEMENT BINDING PROTEIN 3
AT2G4627031872244.05,25527,20619.32.09 × 10–531.82 × 10–51GBF3G-BOX BINDING FACTOR 3
AT1G0106051772271.611,89627,20643.73.01 × 10–532.10 × 10–51LHYLATE ELONGATED HYPOCOTYL
AT2G3627027472238.04,18827,20615.48.13 × 10–514.73 × 10–49ABI5, GIA1GROWTH-INSENSITIVITY TO ABA 1, ABA INSENSITIVE 5
AT3G6242032772245.35,76427,20621.27.54 × 10–493.76 × 10–47BZIP53BASIC REGION/LEUCINE ZIPPER MOTIF 53
AT1G1833061972285.716,87827,20662.03.63 × 10–461.58 × 10–44EPR1, RVE7EARLY-PHYTOCHROME-RESPONSIVE 1, REVEILLE 7
AT5G1730058572281.015,40327,20656.66.78 × 10–452.63 × 10–43RVE1REVEILLE 1
AT1G3215035772249.46,97927,20625.76.05 × 10–442.11 × 10–42bZIP68BASIC REGION/LEUCINE ZIPPER TRANSCRIPTION FACTOR 68
AT4G3459038172252.87,78127,20628.61.94 × 10–436.15 × 10–42GBF6, BZIP11, ATB2ARABIDOPSIS THALIANA BASIC LEUCINE-ZIPPER 11, G-BOX BINDING FACTOR 6
AT5G5266022472231.03,28027,20612.16.20 × 10–431.80 × 10–41
AT5G1583033672246.56,44027,20623.72.94 × 10–427.88 × 10–41bZIP3BASIC LEUCINE-ZIPPER 3
AT2G1816017872224.72,26827,2068.34.60 × 10–411.15 × 10–39GBF5, bZIP2, ATBZIP2, FTM3BASIC LEUCINE-ZIPPER 2, FLORAL TRANSITION AT THE MERISTEM3, G-BOX BINDING FACTOR 5
AT4G0128033972247.06,65427,20624.51.88 × 10–404.38 × 10–39
AT3G1011357972280.215,66427,20657.66.91 × 10–391.51 × 10–37
AT1G4524916572222.92,11227,2067.81.45 × 10–372.98 × 10–36ABF2, AREB1ABSCISIC ACID RESPONSIVE ELEMENTS-BINDING PROTEIN 1, ABSCISIC ACID RESPONSIVE ELEMENTS-BINDING FACTOR 2
AT3G1080013272218.31,46927,2065.47.86 × 10–361.52 × 10–34BZIP28
AT4G3678026972237.34,94427,20618.21.02 × 10–341.88 × 10–33BEH2BES1/BZR1 HOMOLOG 2
AT2G3553013772219.01,63027,2066.03.89 × 10–346.79 × 10–33bZIP16BASIC REGION/LEUCINE ZIPPER TRANSCRIPTION FACTOR 16

Summary statistics of the upstream transcription factor (TF) enrichment analysis for the Arabidopsis gene set bound by LHY ().

aThe number of genes bound by the specific TF in the test set. bThe number of genes in the test set. cThe number of genes bound by the specific TF in the reference set. dThe number of genes in the reference set. eThe p-value after Bonferroni or Benjamini-Hochberg correction.

We also compared EAT-UpTF analysis to a conventional motif enrichment analysis for a similar purpose. DREME, a motif enrichment algorithm of MEME suite (), identified 33 conserved sequence motifs that can be bound by 157 TFs (Supplementary Table 1). While the LHY transcription factor was predicted, which could bind to two motifs, AAATATCK and GATATTTW (Supplementary Table 1), a vast number of additional cis-elements, which are not related to LHY, were also suggested. These results indicate that a motif enrichment analysis possibly produces a considerable number of false positives, but EAT-UpTF enables to suggest realistic upstream TFs.

To ensure whether the EAT-UpTF analysis is relevant with less stringent data set, we input DEGs in ccal lhy double mutant relative to wild type identified by RNA-seq (). Again, EAT-UpTF identified LHY as an over-represented upstream TF for the input gene set (Table 2). Since CCA1 and LHY are transcriptional repressors (), a significant portion of up-regulated genes in cca1 lhy was supposed to be direct targets of CCA1 and LHY. Indeed, EAT-UpTF predicted LHY as a top ranked TF for up-regulated genes in cca1 lhy double mutant (Supplementary Table 2), whereas LHY was excluded but other bZIP TFs were identified to be bound to down-regulated genes in cca1 lhy (Supplementary Table 3).

TABLE 2

TF ID (AGI ID)xanbObserved (%)XcNdExpected (%)p-ValueCorrected p-alueeGene symbolsGene names
AT5G0284026782432.44,11027,20615.19.65 × 10–373.37 × 10–34LCL1LHY/CCA1-LIKE 1
AT4G0128032982439.96,65427,20624.51.75 × 10–233.05 × 10–21
AT5G5266019682423.83,28027,20612.11.71 × 10–211.98 × 10–19
AT3G0960037482445.48,27627,20630.43.27 × 10–202.85 × 10–18LCL5, RVE8LHY-CCA1-LIKE5, REVEILLE 8
AT1G0106047982458.111,89627,20643.72.47 × 10–171.72 × 10–15LHY1, LHYLATE ELONGATED HYPOCOTYL 1, LATE ELONGATED HYPOCOTYL
AT3G6242027582433.45,76427,20621.21.20 × 10–166.99 × 10–15BZIP53BASIC REGION/LEUCINE ZIPPER MOTIF 53
AT4G3459034482441.77,78127,20628.61.75 × 10–168.70 × 10–15BZIP11, GBF6, ATB2G-BOX BINDING FACTOR 6, NA, ARABIDOPSIS THALIANA BASIC LEUCINE-ZIPPER 11
AT2G4627025082430.35,25527,20619.39.54 × 10–154.16 × 10–13GBF3G-BOX BINDING FACTOR 3
AT1G1833061082474.016,87827,20662.09.71 × 10–143.76 × 10–12RVE7, EPR1REVEILLE 7, EARLY-PHYTOCHROME-RESPONSIVE1
AT3G5685019482423.53,93627,20614.51.39 × 10–124.86 × 10–11AREB3, DPBF3ABA-RESPONSIVE ELEMENT BINDING PROTEIN 3
AT5G1730056082468.015,40327,20656.68.36 × 10–122.65 × 10–10RVE1REVEILLE 1
AT1G3215029782436.06,97927,20625.71.37 × 10–113.97 × 10–10bZIP68,BASIC REGION/LEUCINE ZIPPER TRANSCRIPTION FACTOR 68
AT2G1816012682415.32,26827,2068.31.78 × 10–114.77 × 10–10GBF5, bZIP2, FTM3BASIC LEUCINE-ZIPPER 2, G-BOX BINDING FACTOR 5, FLORAL TRANSITION AT THE MERISTEM 3
AT5G1583027882433.76,44027,20623.72.02 × 10–115.03 × 10–10bZIP3BASIC LEUCINE-ZIPPER 3
AT1G4524911982414.42,11227,2067.82.95 × 10–116.87 × 10–10AREB1, ABF2ABSCISIC ACID RESPONSIVE ELEMENTS-BINDING FACTOR 2, ABSCISIC ACID RESPONSIVE ELEMENTS-BINDING PROTEIN 1
AT2G3627019882424.04,18827,20615.43.38 × 10–117.38 × 10–10GIA1, ABI5GROWTH-INSENSITIVITY TO ABA 1, ABA INSENSITIVE 5
AT2G355309782411.81,63027,2066.01.44 × 10–102.96 × 10–9bZIP16,BASIC REGION/LEUCINE ZIPPER TRANSCRIPTION FACTOR 16
AT3G1011355982467.815,66427,20657.65.25 × 10–101.02 × 10–8
AT1G7539012782415.42,48527,2069.12.97 × 10–95.46 × 10–8bZIP44BASIC LEUCINE-ZIPPER 44
AT3G108008282410.01,46927,2065.47.24 × 10–81.26 × 10–6BZIP28

Summary statistics of enriched upstream TFs for differentially expressed genes (DEGs) in cca1lhy double mutant ().

aThe number of genes bound by the specific TF in the test set. bThe number of genes in the test set. cThe number of genes bound by the specific TF in the reference set. dThe number of genes in the reference set. eThe p-value after Bonferroni or Benjamini-Hochberg correction.

In addition, we further examined the relevance of EAT-UpTF in upstream TF enrichment analysis using unoptimized datasets. Genes up-regulated and down-regulated in root tissues upon 1 μM IAA treatment for 6 h () were used as input queries. As for the up-regulated genes, EAT-UpTF identified LATERAL ORGAN BOUNDARIES DOMAIN 19 (LBD19), LBD18 and LBD16 as upstream regulators, which are involved in auxin-dependent lateral root emergence () (Table 3). Meanwhile, BASIC REGION/LEUCINE ZIPPER MOTIF 53 (bZIP53) and bZIP11, which negatively regulate adventitious root formation and primary root growth in an auxin-dependent pathway (; ), were retrieved as overrepresented upstream TFs for the IAA-repressed genes (Table 4). Overall, the EAT-UpTF analysis reliably identified upstream TFs for a group of GOIs. Although our study mainly focused on the enriched upstream TFs for input query genes, which provides essential interpretation of the GOIs in the context of biological pathways and networks, we cannot rule out that TFs regulating a subset of input genes are also sometimes important for estimating biological functions of GOIs, independently of statistical enrichment. Thus, EAT-UpTF can also be used for profiling all possible upstream TFs that potentially regulate GOIs.

TABLE 3

TF ID (AGI ID)xanbObserved (%)XcNdExpected (%)p-ValueCorrected p-valueeGene symbolsGene names
AT1G7274017278921.82,92427,20610.75.21 × 10–201.82 × 10–17
AT2G4541030378938.46,83527,20625.14.78 × 10–171.67 × 10–14LBD19LOB DOMAIN-CONTAINING PROTEIN 19
AT2G4542021578927.24,50327,20616.61.11 × 10–143.88 × 10–12LBD18LOB DOMAIN-CONTAINING PROTEIN 18
AT5G59430497896.256327,2062.19.88 × 10–123.45 × 10–9TRP1,TELOMERIC REPEAT BINDING PROTEIN 1
AT3G46590337894.236327,2061.38.56 × 10–92.99 × 10–6TRP2, TRFL1, ATTRP2TRF-LIKE 1
AT5G6758022178928.05,44627,20620.02.85 × 10–89.94 × 10–6TRB2, TBP3TELOMERE-BINDING PROTEIN 3, TELOMERE REPEAT BINDING FACTOR 2
AT1G3467013678917.23,08627,20611.33.89 × 10–71.36 × 10–4MYB93MYB DOMAIN PROTEIN 93
AT4G3273026978934.17,32227,20626.93.87 × 10–61.35 × 10–3MYB3R1, PC-MYB1MYB DOMAIN PROTEIN 3R1, C-MYB-LIKE TRANSCRIPTION FACTOR 3R-1
AT5G115108378910.51,73227,2066.44.80 × 10–61.68 × 10–3AtMYB3R4MYB DOMAIN PROTEIN 3R4
AT2G0282024978931.66,79427,20625.01.36 × 10–54.75 × 10–3MYB88MYB DOMAIN PROTEIN 88
AT3G10030427895.375127,2062.84.44 × 10–51.55 × 10–2
AT1G0618010278912.92,42227,2068.98.38 × 10–52.93 × 10–2ATMYBLFGN, MYB13MYB DOMAIN PROTEIN 13
AT3G1521017978922.74,75827,20617.59.38 × 10–53.28 × 10–2ERF4, RAP2.5RELATED TO AP2 5, ETHYLENE RESPONSIVE ELEMENT BINDING FACTOR 4
AT3G0407023178929.36,44827,20623.71.49 × 10–45.20 × 10–2NAC047NAC DOMAIN CONTAINING PROTEIN 47
AT5G023209778912.32,33427,2068.62.04 × 10–47.11 × 10–2MYB3R5MYB DOMAIN PROTEIN 3R-5
AT5G5885018178922.94,89527,20618.02.13 × 10–47.44 × 10–2MYB119MYB DOMAIN PROTEIN 119
AT1G2837020578926.05,65727,20620.82.21 × 10–47.72 × 10–2ERF11ERF DOMAIN PROTEIN 11
AT5G2519016178920.44,28127,20615.72.38 × 10–48.32 × 10–2ESE3ETHYLENE AND SALT INDUCIBLE 3
AT5G65130767899.61,74227,2066.42.54 × 10–48.87 × 10–2
AT2G42430317893.954027,2062.02.75 × 10–49.60 × 10–2ASL18, LBD16LATERAL ORGAN BOUNDARIES-DOMAIN 16, ASYMMETRIC LEAVES2-LIKE 18

Summary statistics of enriched upstream TFs for up-regulated genes in Arabidopsis roots upon 1 μM IAA treatment for 6 h ().

aThe number of genes bound by the specific TF in the test set. bThe number of genes in the test set. cThe number of genes bound by the specific TF in the reference set. dThe number of genes in the reference set. eThe p-value after Bonferroni or Benjamini-Hochberg correction.

TABLE 4

TF ID (AGI ID)xanbObserved (%)XcNdExpected (%)p-ValueCorrected p-valueeGene symbolsGene names
AT3G6242023865936.15,76427,20621.23.78 × 10–191.32 × 10–16BZIP53BASIC REGION/LEUCINE ZIPPER MOTIF 53
AT4G3459028965943.97,78127,20628.62.33 × 10–178.12 × 10–15BZIP11, GBF6, bZIP11, ATB2G-BOX BINDING FACTOR 6, BASIC LEUCINE-ZIPPER 11
AT5G6531045165968.414,29527,20652.53.52 × 10–171.23 × 10–14ATHB5,HOMEOBOX PROTEIN 5
AT4G3674046065969.814,74227,20654.28.29 × 10–172.89 × 10–14HB-5, ATHB40HOMEOBOX PROTEIN 40
AT5G6670028365942.97,65827,20628.11.44 × 10–165.03 × 10–14HB-8, ATHB53HOMEOBOX-8, HOMEOBOX 53
AT5G0379038165957.811,60527,20642.71.77 × 10–156.17 × 10–13ATHB51, LMI1HOMEOBOX 51, LATE MERISTEM IDENTITY1
AT5G1583024465937.06,44027,20623.75.41 × 10–151.89 × 10–12bZIP3BASIC LEUCINE-ZIPPER 3
AT1G1468739365959.612,17627,20644.86.05 × 10–152.11 × 10–12HB32, ZHD14HOMEOBOX PROTEIN 32, ZINC FINGER HOMEODOMAIN 14
AT3G5685016965925.63,93627,20614.51.84 × 10–146.42 × 10–12AREB3, DPBF3ABA-RESPONSIVE ELEMENT BINDING PROTEIN 3
AT1G6978042265964.013,48627,20649.62.64 × 10–149.22 × 10–12ATHB13
AT1G1263022865934.65,96027,20621.92.79 × 10–149.72 × 10–12
AT1G3215025465938.56,97927,20625.71.30 × 10–134.53 × 10–11bZIP68BASIC REGION/LEUCINE ZIPPER TRANSCRIPTION FACTOR 68
AT2G1855022965934.76,12427,20622.52.91 × 10–131.01 × 10–10ATHB21, HB-2HOMEOBOX-2, HOMEOBOX PROTEIN 21
AT3G5026040065960.712,82527,20647.11.07 × 10–123.73 × 10–10DEAR1, ATERF#011, CEJ1COOPERATIVELY REGULATED BY ETHYLENE AND JASMONATE 1, DREB AND EAR MOTIF PROTEIN 1
AT2G1816011065916.72,26827,2068.31.48 × 10–125.17 × 10–10bZIP2, FTM3, ATBZIP2, GBF5G-BOX BINDING FACTOR 5, BASIC LEUCINE-ZIPPER 2, FLORAL TRANSITION AT THE MERISTEM3
AT2G4627020165930.55,25527,20619.32.40 × 10–128.38 × 10–10GBF3G-BOX BINDING FACTOR 3
AT5G5202022465934.06,06927,20622.32.49 × 10–128.69 × 10–10
AT4G1675035365953.610,97127,20640.32.63 × 10–129.18 × 10–10
AT1G7539011565917.52,48527,2069.18.58 × 10–123.00 × 10–9bZIP44BASIC LEUCINE-ZIPPER 44
AT1G6901032865949.810,13227,20637.22.20 × 10–117.69 × 10–9BIM2BES1-INTERACTING MYC-LIKE PROTEIN 2
AT2G3627016565925.04,18827,20615.45.50 × 10–111.92 × 10–8ABI5, GIA1ABA INSENSITIVE 5, GROWTH-INSENSITIVITY TO ABA 1
AT5G5199027965942.38,32527,20630.67.80 × 10–112.72 × 10–8DREB1D, CBF4C-REPEAT-BINDING FACTOR 4, DEHYDRATION-RESPONSIVE ELEMENT-BINDING PROTEIN 1D
AT5G258108265912.41,60227,2065.91.22 × 10–104.26 × 10–8TNYTINY
AT1G7145031065947.09,57427,20635.21.60 × 10–105.60 × 10–8
AT3G108007765911.71,46927,2065.41.62 × 10–105.64 × 10–8BZIP28
AT1G7720017565926.64,64627,20617.14.32 × 10–101.51 × 10–7
AT4G2548015865924.04,06427,20614.94.46 × 10–101.56 × 10–7DREB1A, CBF3C-REPEAT BINDING FACTOR 3, DEHYDRATION RESPONSE ELEMENT B1A
AT2G31220556598.391327,2063.46.64 × 10–102.32 × 10–7
AT3G2892020365930.85,71127,20621.01.42 × 10–94.97 × 10–7ZHD9, AtHB34ZINC FINGER HOMEODOMAIN 9, HOMEOBOX PROTEIN 34
AT4G3273024665937.37,32227,20626.92.20 × 10–97.67 × 10–7MYB3R1, PC-MYB1MYB DOMAIN PROTEIN 3R1, C-MYB-LIKE TRANSCRIPTION FACTOR 3R-1

Summary statistics of enriched upstream TFs for down-regulated genes in Arabidopsis roots upon 1μM IAA treatment for 6 h ().

aThe number of genes bound by the specific TF in the test set. bThe number of genes in the test set. cThe number of genes bound by the specific TF in the reference set. dThe number of genes in the reference set. eThe p-value after Bonferroni or Benjamini-Hochberg correction.

The EAT-UpTF analysis requires the input of an experimentally validated genome-wide list of TF-target genes in the form of locus ID. As mentioned above, we used the DAP-seq Arabidopsis database for the initial validation of EAT-UpTF. However, the EAT-UpTF analysis is not limited to the use of DAP-seq data and could also employ ChIP-seq data or any database that provides a list of TF-target genes. If only ‘bed’ files for DAP-seq and ChIP-seq are available, they can be converted to the EAT-upTF input format (Figure 1; see EAT-upTF homepage). In this regard, the EAT-UpTF analysis could be expanded to any plant species for which DAP-seq, ChIP-seq, or other appropriate sequencing data are available. In the future, a large-scale database integrating DAP-seq and ChIP-seq results would aid the identification of bona fide upstream TFs for groups of GOIs. EAT-UpTF is an open platform that can be improved by integrating updated TF databases. In addition, to ensure convenience for users, TF regulatory networks of GOIs identified by EAT-UpTF can also be visualized in Cytoscape (Figure 2). Compared to the previous webtools, such as TF2Network () and AthaMap (), which conduct cis-element-based construction of TF regulatory networks, EAT-UpTF involves simple and rapid processing of data without cis-element identification, and thereby promptly visualizes gene regulatory networks showing TF-target gene interactions. While processing our study, a remarkable webtool ‘Plant Regulomics’2 has been released (), which might implement a similar logic and code of EAT-UpTF, supporting the relevance of this analysis.

FIGURE 2

) was used as a test input. The area of a node represents the edge count and the color intensity indicates the strength of the neighborhood connectivity. Black dots represent single nodes.

Conclusion

In summary, EAT-UpTF is a tool for analyzing the over-representation of upstream TFs based on the relative enrichment of TF-target genes in a group of GOIs in plants. EAT-UpTF can be used to identify upstream TFs for a group of genes without limitations on species and conservation of cis-motifs. With a regular update or manual construction of databases of TF-target genes in plant species, EAT-UpTF will become a powerful tool for TF regulatory network studies in plants. For user convenience, EAT-UpTF web service is also available at http://chromatindynamics.snu.ac.kr:8080/EatupTF.

Statements

Data availability statement

EAT-UpTF is available at https://github.com/sangreashim/EAT-UpTF; operating system(s): Linux, programming languages: Python3; other requirements: Python3 packages (SciPy, Statsmodels, Argparse). The EAT-UpTF home page provides detailed user manuals. EAT-UpTF is freely available. There are no restrictions on non-academics use.

Author contributions

SS and PS: conceptualization and funding acquisition. SS: data curation and implementation and writing – original draft. PS: writing – review and editing. Both authors contributed to the article and approved the submitted version.

Funding

This work was supported by the National Research Foundation of Korea (NRF-2019R1I1A1A01061376 to SS, NRF-2019R1A2C2006915 to PS) and the Rural Development Administration (PJ01319304 to PS).

Acknowledgments

This manuscript has been released as a pre-print at bioRXiv (doi: https://doi.org/10.1101/2020.03.22.001537) (SS and PS).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene.2020.566569/full#supplementary-material

Footnotes

1.^http://neomorph.salk.edu/dap_web/pages/index.php

2.^http://bioinfo.sibs.ac.cn/plant-regulomics/

References

Summary

Keywords

transcription factor, cis-elements, plant, Arabidopsis, DAP-seq

Citation

Shim S and Seo PJ (2020) EAT-UpTF: Enrichment Analysis Tool for Upstream Transcription Factors of a Group of Plant Genes. Front. Genet. 11:566569. doi: 10.3389/fgene.2020.566569

Received

02 June 2020

Accepted

17 August 2020

Published

11 September 2020

Volume

11 - 2020

Edited by

Nunzio D’Agostino, University of Naples Federico II, Italy

Reviewed by

Federico Zambelli, University of Milan, Italy; Jose M. Franco-Zorrilla, National Center for Biotechnology (CNB), Spain; Yang Jae Kang, Gyeongsang National University, South Korea

Updates

Copyright

© 2020 Shim and Seo.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Sangrea Shim, Pil Joon Seo,

This article was submitted to Computational Genomics, a section of the journal Frontiers in Genetics

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

Outline

Figures

Cite article

Copy to clipboard


Export citation file


Share article

Share on WeChat

Scan with WeChat to share this article

Article metrics