About


PlanNET is a web application that stores predicted Schmidtea mediterranea protein-protein interactions (ppi) projected over a Human ppi network. Currently, PlanNET holds 273,203 human ppi, and 729,043 Schmidtea mediterranea interactions (predicted for 9 different transcriptomes).

PlanNET has been developed by the Computational Genomics Lab, at the University of Barcelona.



Citing PlanNET


PlanNET: homology-based predicted interactome for multiple planarian transcriptomes
S. Castillo-Lara, J.F. Abril
Bioinformatics, Volume 34, Issue 6, 15 March 2018, Pages 1016–1023, btx738

BibTex
    @article{Castillo-Lara2018,
      author = {Castillo-Lara, S and Abril, J F},
      title = {"PlanNET: homology-based predicted interactome for multiple planarian transcriptomes"},
      journal = {Bioinformatics},
      volume = {34},
      number = {6},
      pages = {1016--1023},
      year = {2018},
      doi = {10.1093/bioinformatics/btx738},
      URL = {http://dx.doi.org/10.1093/bioinformatics/btx738},
      eprint = {/oup/backfile/content_public/journal/bioinformatics/34/6/10.1093_bioinformatics_btx738/2/btx738.pdf}
    }
        


About the PPI predictor


PlanNET uses a Random Forest classifier in order to predict protein-protein interactions for the planaria Schmidtea mediterranea, by using a reference human network. This predictor takes into account 19 features for each possible pair of interacting proteins, and then it assigns a score to each pair, ranging from 0 to 1.

The prediction of interactions was evaluated using OOB values of performance. The performance measures for the current version of PlanNET are:

Precision Recall Specificity F-measure Accuracy
31.23 % 27.80 % 99.75 % 29.42 % 99.46 %
ROC curve of the random forest classifiers built using the OOB votes of the predictor.
Variable importance of the 19 features used for the classification expressed as the mean decrease in Gini coefficient.

For more information refer to the published paper.



Update Notes


v1.2: January 23rd 2019

  • Schmidtea mediterranea predicted genes (see: A. Rozanski et al.) as well as their associated transcripts have been included in PlanNET as SmesGene and Smest respectively.
  • Smest gene associated transcripts are now available in the BLAST application.
  • A new gene card for Smesgenes has been implemented, with information regarding the associated transcripts across different transcriptomes, the genomic location, gene ontologies and predicted protein-protein interactions.
  • New Quick Search on Navigation bar has been added, along with a search All Databases option for GeneSearch.
  • Ability to filter results on GeneSearch implemented.
  • Several genomic features (such as mapped transcripts over the newly assembled S.med genome) can now be visualized using our new JBrowse plugin.
  • PlanExp, a PlanNET plugin for exploring RNA-seq experiments is available for registered users to test.
  • Databases are now identified by node border color instead of shape.
  • Node shape in NetExplorer and PathwayFinder now refers to the technology used for sequencing, with triangles reserved for "legacy" technologies such as Sanger and 454, and circles for high-throughput RNA-seq.
  • General improvements to design and readability of the application.


v1.1: August 7th 2018

PlanNET has been updated from version 1.0 (published version at Bioinformatics) to version 1.1. These are the most important changes:
  • In order to look for homologous proteins for each planarian transcript, we prioritized BLAST instead of EggNOG alignments.
  • The RandomForest classifier was refined by filtering the training dataset further, removing interactions not annotated as physical from BioGRID.
  • We selected a score cut-off of 0.7 for interactions, instead of the previously used 0.6 cut-off. The Filter Interactions slider in NetExplorer was updated: now it only shows interactions with a score higher than 0.7 by default, but the user can still select other cut-off values
  • PlanNET now holds 1,360,991 Schmidtea mediterranea interactions (up from 510,996).
  • The Smed454 dataset and Newmark dataset were temporarily removed because of disk size limitations. These will be uploaded in due time.
  • A manually curated list of identifiers for planarian transcripts was uploaded to PlanNET. Now users can search for these identifiers, which follow the form: "Smed_XXX". If a particular transcripts has a manually curated identifier, the identifier will appear in pink both in Gene Search and in the corresponding Gene Card.
  • New Data Downloader allows users to download sequences and sequence annotations in bulk. Available at downloads page.
  • Many quality of life improvements to the overall design of the website, including a search bar in NetExplorer.