cropPAL²⁰²⁰ species annotation information

cropPAL²⁰²⁰ is updated about once a year, which means experimental and computed data increase with every update. The current version cropPAL²⁰²⁰ is built on the Ensembl Pantmart version 40 described in the Gramene release 58 notes. The version 2 of cropPAL expanded the curated experimental and precomputed data to 12 species. The data was linked to the proteome annotations listed below.

cropPAL²⁰²⁰ experimental data

An overview of the experimental studies in cropPAL²⁰²⁰ as of October 2019 is shown below. Studies published after this date will be included in the next update. The experimental data is captured and obsolete or non-Ensemble Plant protein IDs are either cross-references or sequences belonging to these IDs are retrieved and BLASTed against the current proteome. This helps retain valuable experimental data and links it to current genome standards. The table below shows the number of studies that have been linked to the Ensembl proteomes as well as the number of PPI studies that were derived from Arabidopsis through homology linking.

The table below details the number of experimental localisations, the number of distinct proteins localised and available PPI pairs infered from Arabidopsis by homology linking for each species.

cropPAL²⁰²⁰ predictors

There are 12 predictors integrated into cropPAL which use distinct training data sets, input variables and prediction methods. These have been reviewed and compared for their contribution to the SUBA consensus call in our recent study about SUBAcon. Predictors vary in their accuracy for each subcellular compartment. Information describing the performance of individual predictors in specific subcellular compartments in Arabidopsis can be found on the SUBA4 about page . For 6 predictors (MultiLoc2, predotar, targetP, YLoc, iPSORT, WolfPSORT) proteome-wide prediction sets exists for all cropPAL²⁰²⁰ species. The remaining 6 predictors (BaCeLo, PProwler, ChloroP, EpiLOC, PTS1, Plant-mPLoc) were either not accessible or too slow to complete the proteome at time of release. Incomplete prediction sets and predictions from previous cropPAL versions for identical protein sequences are provided for these algorithms in cropPAL²⁰²⁰. Additional predictions will be added at the next update.

cropPAL²⁰²⁰ Winner-takes-all locations

The winner-takes-all (WTA) is a calculated output that attempts to unify prediction as well as available experimental data. This is different to a classifier that is based on a training set and tested performance. The WTA call is generated by counting up the predicted locations as 1 vote (e.g. 2 x mitochondrial predictions = 1 vote for mitochondrion) and adding any experimental verification as 1 vote each (e.g. 2 x plastidal GFP localisations = 2 votes for plastid). The votes are added up and the winner(s) is the exclusive location suggestion for this protein. In the example, the votes would result in a 2:1 plastidal output with a final call to plastid. This strategies ensures a stronger influence of experimental verifications when they are available. In case of several locations with equal votes as the winner all winners will be chosen. To avoid skewing of the calculations only complete proteome predictions data sets could be used for this calculation. This included the predictors: MultiLoc2, predotar, targetP, YLoc, iPSORT, WolfPSORT.

cropPAL²⁰²⁰ Protein-Protein-Interaction Data

Direct experimental evidence of Protein-Protein Interaction data for crop species is still sparse. For Arabidopsis, SUBA4 collated over 26000 experiments that showed the interaction of two or more proteins covering a third of the proteome. The similarity in protein sequence of Arabidopsis PPI pairs across species may help find PPI partners in crop species for hypothesis-driven research. In cropPAL²⁰²⁰, we have taken all evidence-based Arabidopsis PPI data and linked it to homologically similar proteins in crop species. This yielded over 800000 suggested PPI partners across the 12 crop species. The homology linking was performed using TreeBEST homology linking as available through Ensembl Plant Mart version 40.

cropPAL2 Homology Linking

There are two strategies in cropPAL²⁰²⁰ linking crop data to other species. One apporach that is used to determine the best match is a reciprocal BLAST of all crop proteins against Arabidopsis. This generates a link to the nearest match and description in SUBA4.
The second type of homology linking is performed using the Ensembl Plants gene tree as provided by Gramene. The tree was generated using TreeBeST and was described in more detail in 2016 in DataBase. The TreeBeST homology linking links all cropPAL²⁰²⁰ species to each other as well as to Arabidopsis. The gene tree is a more conservative measure than the reciprocal blast

Site Data

The data backing this website is fully accessible via the OpenAPI Specification (OAS). You can view and download data now using our own page here. Or you can inspect and edit the specification in the offsite Swagger editor.