MyMiner project

Objectives.

The goal of this application is to provide tools and methods for researchers to easily and rapidly classify scientific documents or biomedical terms.

This application has been developed with the following ambitious goals :
  1. To speed up the process of information transfer from scientific literature to computationally exploitable data.
  2. To facilitate the comparison and the evaluation of dataset sets of biological documents. The definition of high quality reference datasets will in turn improve machine learning technique predictions.
  3. To allow researchers to define topic-specific databases.
  4. To serve, promote and facilitate community annotation efforts.

MyMiner application integrates several distinct tools and options which contributes to a better and easier classification.

Use.

We utilised this application to create a Muscle specific database called MyoBase (in development). To create a "muscle" dedicated system, we have tagged biological entities to determine which genes, diseases and GO terms are related to muscle development, function and repair from the generalist and cross species comparison system Compare. We also used the MyMiner to train and test a semiu-supervised machine learning program to classify articles according to the muscle topic.

Teams involved.

This application is the result of a collaborative work between Martin Krallinger from the CNIO (Centro Nacional de Investigaciones Oncologicas) Madrid SPAIN and David Salgado from the IBDML (The Developmental Biology Institute of Marseilles Luminy) Marseilles FRANCE and now from ARMI (Australian Regenerative Medicine Institute) at Monash University Melbourne AUSTRALIA.

With the participation of :
Marc Depaule (IBDML).
Elodie Drula (IBDML).
Ashish Tendulkar (CNIO).

And the supervision of :
Alfonso Valencia (CNIO) and Christophe Marcelle(ARMI).

Global view

Schematic representation of the MyMiner application.

Click on the picture to zoom in. MyMiner Global View


For a better experience with MyMiner, we recommend using Firefox or Chrome.