Maize Interactome

1 HomePage

The home page has three sections: 1. An overview of the database; 2. The quick link to important functions; 3. The Interactome Network Summary Information.


2 Network Display

We design different levels of networks to display our data. Firstly, according to the user usage, it is divided into three types: Hub Network, Directly Connected Network, and Module Network. Users utilize them to access the different levels of data. Secondly, based on the data source, the network can be divided into 3D Network, Transcriptome Network, Translation Network, PPI Network, and Integrative Omics Network. The Transcriptome Network and Integrative Omics Network could separate into complete networks, slim networks (based on B73V4 annotation), and 'without siRNAs network' that remove large amounts of unknown functional siRNAs. In addition, each network has different confidence levels (high, middle, and low). In short, we built 25 networks based on the above content.

2.1 Hub network

Hub network is the essence of the network. We construct the network by screening the top 10% nodes in 100 to 200 modules to overview the network.
1. The red box 1 represents the omics and confidence of the network.
2. The red box 2 indicates the level of the network (Hub Network, Directly Connected Network, and Module Network).
3. In the red box 3, users can search for Directly Connected Network and Module Network in different confidences and omics.
4. In the red box 4, you can search for nodes in the current network and change the background color (Light Color and Dark Color).


2.2 Directly Connected Network

A node's directly connected network is formed based on the interaction between directly connected nodes. It is mainly formed by the direct interaction of the node and the interaction between the interacting nodes.


2.3 Module Network

Sometimes, a node's directly connected network is difficult to help biologists understanding the role of a node in the network. Therefore, the Module network is used to portray the role of nodes in the network. Furthermore, the module network could monitor the target node's connection farther and more critically under limited resources.



2.4 3D Network

3D network from the ChIA-PET experiments of B73.


2.5 Transcriptome Network

Transcriptome without siRNAs Network:Transcriptome without siRNAs co-expression network is constructed by the RNA expressions (including microRNA, circle RNA, fusion RNA, long non-coding RNA, and mRNA loci) from 26 tissues of B73. The division of confidence is assigned based on a curve graph in which the total node numbers are below a certain MR value. The high confidence level is at the inflection point, the middle confidence level is at the flat position, and the low confidence level is at the completely flat position.
Slim-Transcriptome Network: Transcriptome co-expression network is constructed using the gene loci annotated by B73 version4 file from 26 tissues of B73.
Transcriptome Network: Transcriptome co-expression network is constructed by the RNA expressions ( including sRNA, circle RNA, fusion RNA, long non-coding RNA, and mRNA loci) from 26 tissues of B73.

Transctiptome without siRNAs Network (low confidence)


Transctiptome without siRNAs Network (middle confidence)


Transctiptome without siRNAs Network (high confidence)


Slim-Transcriptome Network (low confidence)


Slim-Transcriptome Network (middle confidence)


Slim-Transcriptome Network (high confidence)


Transcriptome network (low confidence)


Transcriptome network (middle confidence)


Transcriptome network (high confidence)


2.6 Translatome Network

The translatome network is constructed by the ribosome-encapsulated RNA expressions level from 20 tissues of B73. The division of confidence is assigned based on a curve graph in which the total node numbers are below a certain MR value. The high confidence level is at the inflection point, the middle confidence level is at the flat position, and the low confidence level is at the completely flat position.

Translatome Network (low confidence)


Translatome Network (middle confidence)


Translatome Network (high confidence)


2.7 PPI Network

Protein-protein interactions network is constructed by high-throughput Yeast-2-hybridization experiments across eight distinct tissues of B73. The low confidence include all protein-protein interactions we tested, the middle confidence is the protein-protein interactions removing the no-load experiments. And the high confidence in protein-protein interactions when using the no-load experiments as a positive control, the possible self-activation interaction pairs are removed by machine learning.

PPI Network (low confidence)


PPI Network (middle confidence)


PPI Network (high confidence)


2.8 Integrative Omics Network

Integrative Omics without siRNAs Network consists of corresponding 3D, transcriptome without siRNA, translatome, and PPIs datasets. The edge score is cubed. If two edges are present, the product of the two edges is multiplied by the average of these two edges (Walley et al., 2019). The three confidence corresponds to the confidence of different networks.
Slim-integrative Omics Network consists of corresponding 3D, Slim-transcriptome, translatome, and PPIs datasets. The three confidence corresponds to the confidence of different networks.
Integrative Omics Network consists of corresponding 3D, transcriptome, translatome, and PPIs datasets. The three confidence corresponds to the confidence of different networks.

Integrative Omics without siRNAs Network (low confidence)


Integrative Omics without siRNAs Network (middle confidence)


Integrative Omics without siRNAs Network (high confidence)


Slim-integrative Omics Network (low confidence)


Slim-integrative Omics Network (middle confidence)


Slim-integrative Omics Network (high confidence)


Integrative Omics Network (low confidence)


Integrative Omics Networkm (middle confidence)


Integrative Omics Network (high confidence)


3 Analysis

We developed four analysis based on our database, namely Differential Gene Network, QTG Interaction Score, Trait Decoder, and Pathway Mapping. These analysis will offer some valuable information to help biologists in identification of the major gene in QTL, differential gene network, biological pathway, and gene regulation of specific traits.

3.1 Differential Gene Network

The user searches for Differential Gene Network by a group of differentially expressed genes. Then the website calculates the average of the shortest distances between the genes provided and divided them into modules.



3.2 QTG Interaction Score

We developed a procedure (QTG Interaction Score) to assist in the localization of QTL candidate genes for exploring unknown genes. The traditional QTL positioning usually builds a multi-generation group to screen and reorganize individual plants to narrow the interval continuously. However, QTG Interaction Score can quickly reduce the range of genes in QTL by the network data. Users can input the QTL position and potential QTL-related Elements to predict the major-effect locus in the QTL position.



3.3 Trait Decoder

The Trait Decoder can predict trait-related genes by some known genes. A machine learning classifier can be used to test different parameters of the selected model by calculating the AUC score and can be predicted whether directly connected genes of the positive sample belong to the trait.



3.4 Pathway Mapping

Pathway Mapping is designed to explore the pathway interaction from different omics.



4 Tools

We develop eight essential tools for users to explore our network data: BLAST, GBrowse, Element Fetch, Element Search, Module Search, Module Hit, Network Creation, and Network Comparison. These tools allow users to access our network data as quickly as possible.

4.1 BLAST

A known sequence is used to obtain the position of the Element ID alignment on the genome.


4.2 GBrowse

The GBrowse can be used to obtain the position on the genome of genomic elements.


4.3 Element Fetch

The users can provide a genomic region to get directly connected nodes in the network or the Element ID lists and provide an Element ID list to get directly connected nodes. The result is in tabular form (such as txt, CSV, excel, and XML downloads).



4.4 Element Search

First, select the type of Element you are searching for. Then enter the searched Element ID, the microRNA module, and the gene module support the ID of the corn B73V4 version. The users can get the result including RNA-seq, Ribo-seq expression, and functional annotations, such as KO and GO. And it shows the module network and the directly connected network. Moreover, the result also provides download edges information of the network.



4.5 Module Search

By Returning the network of the searched module based on our module name, users can download module nodes and edge data in the results page. And these data can be applied to gephi to redraw network distribution.



4.6 Module Hit

Use Fisher's test to determine if the provided Element ID is enriched in the associated module. To avoid enrichment in a large number of small modules, users can set the minimum number of hits to this module node. The default value is 3 (the minimum module in the network is five nodes)




4.7 Network Creation

Create a directly connected network by providing Element IDs. Users can use the space bar or tab to search for the required ID or use a new line to split. The search return page gets a de-redundant network that directly uses the ID to connect to the network. Note that as the number of point input increases, the calculation of the network layout will also increase, and the computing resources consumed will be more.



4.8 Network Comparison

Compare the common and unique parts of the directly connected network formed by the two sets of Element IDs. As with Network Creation, try not to enter too many nodes. The result is a CSV file that provides node and edge information. The node file indicates whether the network is common or unique.



5 Q&A

5.1 What is the naming rule for Element ID?


The naming rule is based on the version4 annotated file. For the mRNA locus annotation, if it is a V4 locus or has a large overlap, the V4 ID is used. If two or more V4 genes are crossed, it is considered to be a new gene. Named according to ZmmRNA000001, ZmmRNA000002.... For lncRNA, if V4 annotates it, then uses the V4 number, while if it is a new site, then ZmLnc000001, ZmLnc000002... will be named; all circular transcript, sRNA, fusion transcript are new comment numbers, named ZmCirc000001, ZmCirc000002...; ZmsRNA000001, ZmsRNA000002...; ZmFusionR000001, ZmFusionR000002... respectively.



5.2 How to choose the type of network when analyzing data?


The 3D data was obtained from the chia-pet experiment, so the interaction pairs obtained were usually close to each other in the genome; the transcriptome and the translatome were constructed based on the correlation of the expression levels of more than 20 tissues; PPI is the physical interaction detected by yeast double hybrids. The edge of the integrated network is at least one omics that has that edge.
Transcriptome network contains the complete component interaction information. Due to the excessive amount of sRNA (~14w), it may affect the network structure. So we constructed the transcriptome without siRNAs, in this network replacing sRNA with microRNA precursor annotated in version4 annotated file. This network is effective for discovering the potential functions of non-coding RNA. All the nodes in the slim-transcriptome network are annotated by version4 annotated file. It is more suitable for the GO enrichment analysis based on annotated genes.



5.3 How to complete an element network-based functional prediction?


For predicting the function of an element component, such as lncRNA, circRNA, etc., you can first retrieve the directly connected network and then do GO or KEGG enrichment analysis to see the possible functions.



5.4 How to choose confidence?


The high, middle, and low confidence have been subjected to a certain threshold screening, and the low confidence network includes all the high confidence edge interactions.