SHREC
2018: Special track on protein shape retrieval
SHREC
2018: Special track on protein shape retrieval
This track is submitted by the structural biology group of the GBA lab (GBA Lab), from the Conservatoire National des Arts-et-Métiers (CNAM, Paris, France), as a section of the SHREC 3D shape retrieval contest 2018.
Proteins are complex macro-molecular non-rigid shapes that can be represented either as a cloud of points or a high-resolution surface. Usually molecular surfaces represent the protein solvent-excluded surface as defined by Connoly (Connoly et al, J Appl Cryst. 1983). Since proteins display multiple possible conformations in solution, the detection of shape similarity and/or identity is of biological relevance in drug discovery processes and molecular characterization of diseases. The proposed dataset contains 2267 individual protein structures, extracted from Protein Data Bank entries. The selected structures display similar size and number of points. The aim of this contest is to retrieve, for each member of a given class, its other conformations from the whole dataset. Both representations (the .pdb format file describing the 3D position of each atom and object format file .off describing the protein molecular surface) of the proteins will be included in the dataset and provided for convenience.
Examples of protein shapes included into the dataset
Aim
* Provide a reliable set of protein molecular surfaces for molecular shape retrieval
* Evaluate state of the art algorithms performance on biological, non-rigid shape retrieval
* Attract experts in the 3D/Vision/shape retrieval fields into applying their methods on molecular shapes
Target Dataset
PDB protein structures are molecular arrangements of one or several protein chains, each chain including one or several domains. The target dataset is built with different domains grouped in classes according to the SCOPe v2.06 database. For each protein, one to fifty conformer structures (non rigid transformation of the shape) issued from Nuclear Magnetic Resonance (NMR) data (i.e. experimentally observed conformations of the protein) were included into the target dataset, resulting in a dataset of 2267 unique protein structure files sharing various levels of surficial similarity. All structures were protonated and solvent-excluded surfaces were calculated using EDTSurf (Xu et al, Plos One 2009).
10 conformers of PDB structure 1NMV and their respective surfaces
PDB/OFF files can be downloaded here:
shape_ retrieval_shrec2018_off_files. tar.gz
shape_ retrieval_shrec2018_pdb_files. tar.gz
Query dataset
The query dataset is the same as the target dataset, the idea is to retrieve for each member of a given class, the other members of the class. Up to 50 conformations are available for a given domain in the dataset.
Ground truth
For each entry of the query dataset, the ground truth is generated by isolating all conformations of the same domain among the target dataset. All information is derived directly from experimental data (NMR)
Evaluation
Standard metrics of previous shape retrieval experiments will be used: precision - recall evaluation, ROC and enrichment curves, Nearest Neighbor, first-tier and second-tier. The participants are expected to return their results as distance matrix file http://shape.cs.princeton.edu/benchmark/documentation/matrix_format.html
Expected number of participants
All 3DOR experts interested into treating non-conventional shapes with inherent complexity such as molecular shapes could be interested. To render the track easily accessible to most participants, we also provide the meshes of the protein Solvent excluded surfaces in the .off format.
Schedule
January 25 - Tha dataset is made available on shrec2018.drugdesign.fr
February 1 - Registration deadline. Registration must be sent to Matthieu Montès and Florent Langenfeld .
February 12 - Extended registration deadline.
February 21 - Submission deadline of the results. Results are submitted along with a one-page description of the method(s) used to generate the results.
February 28 - Track report submission to SHREC for review
March 3 - Reviews are done, notification to the participants.
March 10 - Camera-ready track paper is submitted for inclusion into the proceedings.
April 15-16 - Eurographics Workshop on 3D Object Retrieval 2018, featuring SHREC 2018
Organizers
Matthieu Montès - Conservatoire National des Arts-et-Métiers
Florent Langenfeld - Conservatoire National des Arts-et-Métiers