Limits...
Predicted protein-protein interactions in the moss Physcomitrella patens: a new bioinformatic resource.

Schuette S, Piatkowski B, Corley A, Lang D, Geisler M - BMC Bioinformatics (2015)

Bottom Line: This method has been used to successfully predict interactions for a number of angiosperm plants.Most conserved interactions among proteins were those associated with metabolic processes.Included with this dataset is a method for characterizing subnetworks and investigating specific processes, such as the Calvin-Benson-Bassham cycle.

View Article: PubMed Central - PubMed

Affiliation: Department of Plant Biology, Southern Illinois University, Carbondale, IL, USA. swschuette@gmail.com.

ABSTRACT

Background: Physcomitrella patens, a haploid dominant plant, is fast becoming a useful molecular genetics and bioinformatics tool due to its key phylogenetic position as a bryophyte in the post-genomic era. Genome sequences from select reference species were compared bioinformatically to Physcomitrella patens using reciprocal blasts with the InParanoid software package. A reference protein interaction database assembled using MySQL by compiling BioGrid, BIND, DIP, and Intact databases was queried for moss orthologs existing for both interacting partners. This method has been used to successfully predict interactions for a number of angiosperm plants.

Results: The first predicted protein-protein interactome for a bryophyte based on the interolog method contains 67,740 unique interactions from 5,695 different Physcomitrella patens proteins. Most conserved interactions among proteins were those associated with metabolic processes. Over-represented Gene Ontology categories are reported here.

Conclusion: Addition of moss, a plant representative 200 million years diverged from angiosperms to interactomic research greatly expands the possibility of conducting comparative analyses giving tremendous insight into network evolution of land plants. This work helps demonstrate the utility of "guilt-by-association" models for predicting protein interactions, providing provisional roadmaps that can be explored using experimental approaches. Included with this dataset is a method for characterizing subnetworks and investigating specific processes, such as the Calvin-Benson-Bassham cycle.

Show MeSH
ThePhyscomitrella patensnetwork was viewed in Cytoscape and the hub distribution of proteins in the network were analysed. (A) Large hairy ball of the 67,740 non-redundant interactions in the organic view of Cytoscape. (B) Detailed view showing layers of interacting proteins. (C) Pie chart of hub distribution by node class; “Free ends” are defined as have a single interaction, “pipes” have two interactions, and grouped hubs based on numbers of interactions. Numbers in parentheses indicate number of interacting proteins used to determine hub size.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4384322&req=5

Fig2: ThePhyscomitrella patensnetwork was viewed in Cytoscape and the hub distribution of proteins in the network were analysed. (A) Large hairy ball of the 67,740 non-redundant interactions in the organic view of Cytoscape. (B) Detailed view showing layers of interacting proteins. (C) Pie chart of hub distribution by node class; “Free ends” are defined as have a single interaction, “pipes” have two interactions, and grouped hubs based on numbers of interactions. Numbers in parentheses indicate number of interacting proteins used to determine hub size.

Mentions: More than 104,000 total interactions were predicted from different references including 67,740 unique interactions from 5,694 different Physcomitrella patens proteins (Additional file 2: Table S1). When visualized in its entirety in Cytoscape v3.2.0, the interactome looks like a ball of densely tangled circles and lines (Figure 2a,b). Like a map of roads and towns, this interactome is most useful when plotting connections between genes or pathways of interest and their neighbours, which is made possible in Cytoscape by the select node and zoom tools. Individual interactions were systematically evaluated using an evidence-based confidence value to assess the quality of the predicted interactions. 3936 high confidence (CV > 10), 10,318 medium confidence (CV between 2 and 10), and 53,400 low confidence (CV = 1) interactions were found in moss. The prediction efficiency of different confidence levels was evaluated by comparing the current experimentally determined interactions accumulated at the BioArray Resource [32] to the predicted interactions using the same interlog method and CV calculation for Arabidopsis by Geisler-Lee et al. [11]. The BAR resource combined the work of several high throughput experimental studies [8,33] with individual experiments culled from over 1190 publications (full list available at http://bar.utoronto.ca/interactions/cgi-bin/arabidopsis_interactions_viewer.cgi). The 37645 experimentally determined interactions in Arabidopsis were compared to the 72266 predicted interactions in Arabidopsis from non-plant reference genomes, and 1450 matched. The expected overlap between these datasets by chance alone (equal number of random protein pairs) was 91. This did not account for non-overlapping protein sets (proteins present in one set but absent in the other), nor did it account for the non-overlapping bias, as experimental interactomes tend to be biased towards plant specific genes, of which none would have occurred in the Arabidopsis predicted interactome. When broken down by confidence value, there was only slight improvement comparing interactions with low CV to those with medium CV, and there was 2.1 fold enrichment when comparing interactions with low and high confidence values.Figure 2


Predicted protein-protein interactions in the moss Physcomitrella patens: a new bioinformatic resource.

Schuette S, Piatkowski B, Corley A, Lang D, Geisler M - BMC Bioinformatics (2015)

ThePhyscomitrella patensnetwork was viewed in Cytoscape and the hub distribution of proteins in the network were analysed. (A) Large hairy ball of the 67,740 non-redundant interactions in the organic view of Cytoscape. (B) Detailed view showing layers of interacting proteins. (C) Pie chart of hub distribution by node class; “Free ends” are defined as have a single interaction, “pipes” have two interactions, and grouped hubs based on numbers of interactions. Numbers in parentheses indicate number of interacting proteins used to determine hub size.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4384322&req=5

Fig2: ThePhyscomitrella patensnetwork was viewed in Cytoscape and the hub distribution of proteins in the network were analysed. (A) Large hairy ball of the 67,740 non-redundant interactions in the organic view of Cytoscape. (B) Detailed view showing layers of interacting proteins. (C) Pie chart of hub distribution by node class; “Free ends” are defined as have a single interaction, “pipes” have two interactions, and grouped hubs based on numbers of interactions. Numbers in parentheses indicate number of interacting proteins used to determine hub size.
Mentions: More than 104,000 total interactions were predicted from different references including 67,740 unique interactions from 5,694 different Physcomitrella patens proteins (Additional file 2: Table S1). When visualized in its entirety in Cytoscape v3.2.0, the interactome looks like a ball of densely tangled circles and lines (Figure 2a,b). Like a map of roads and towns, this interactome is most useful when plotting connections between genes or pathways of interest and their neighbours, which is made possible in Cytoscape by the select node and zoom tools. Individual interactions were systematically evaluated using an evidence-based confidence value to assess the quality of the predicted interactions. 3936 high confidence (CV > 10), 10,318 medium confidence (CV between 2 and 10), and 53,400 low confidence (CV = 1) interactions were found in moss. The prediction efficiency of different confidence levels was evaluated by comparing the current experimentally determined interactions accumulated at the BioArray Resource [32] to the predicted interactions using the same interlog method and CV calculation for Arabidopsis by Geisler-Lee et al. [11]. The BAR resource combined the work of several high throughput experimental studies [8,33] with individual experiments culled from over 1190 publications (full list available at http://bar.utoronto.ca/interactions/cgi-bin/arabidopsis_interactions_viewer.cgi). The 37645 experimentally determined interactions in Arabidopsis were compared to the 72266 predicted interactions in Arabidopsis from non-plant reference genomes, and 1450 matched. The expected overlap between these datasets by chance alone (equal number of random protein pairs) was 91. This did not account for non-overlapping protein sets (proteins present in one set but absent in the other), nor did it account for the non-overlapping bias, as experimental interactomes tend to be biased towards plant specific genes, of which none would have occurred in the Arabidopsis predicted interactome. When broken down by confidence value, there was only slight improvement comparing interactions with low CV to those with medium CV, and there was 2.1 fold enrichment when comparing interactions with low and high confidence values.Figure 2

Bottom Line: This method has been used to successfully predict interactions for a number of angiosperm plants.Most conserved interactions among proteins were those associated with metabolic processes.Included with this dataset is a method for characterizing subnetworks and investigating specific processes, such as the Calvin-Benson-Bassham cycle.

View Article: PubMed Central - PubMed

Affiliation: Department of Plant Biology, Southern Illinois University, Carbondale, IL, USA. swschuette@gmail.com.

ABSTRACT

Background: Physcomitrella patens, a haploid dominant plant, is fast becoming a useful molecular genetics and bioinformatics tool due to its key phylogenetic position as a bryophyte in the post-genomic era. Genome sequences from select reference species were compared bioinformatically to Physcomitrella patens using reciprocal blasts with the InParanoid software package. A reference protein interaction database assembled using MySQL by compiling BioGrid, BIND, DIP, and Intact databases was queried for moss orthologs existing for both interacting partners. This method has been used to successfully predict interactions for a number of angiosperm plants.

Results: The first predicted protein-protein interactome for a bryophyte based on the interolog method contains 67,740 unique interactions from 5,695 different Physcomitrella patens proteins. Most conserved interactions among proteins were those associated with metabolic processes. Over-represented Gene Ontology categories are reported here.

Conclusion: Addition of moss, a plant representative 200 million years diverged from angiosperms to interactomic research greatly expands the possibility of conducting comparative analyses giving tremendous insight into network evolution of land plants. This work helps demonstrate the utility of "guilt-by-association" models for predicting protein interactions, providing provisional roadmaps that can be explored using experimental approaches. Included with this dataset is a method for characterizing subnetworks and investigating specific processes, such as the Calvin-Benson-Bassham cycle.

Show MeSH