In order to find out which genes are responsible for diseases such as cancer or diabetes, scientists nowadays frequently resort to using machine-learning models.

Trained models freely available

«What makes Kipoi special is that it provides free access to machine learning models that have already been trained,» says Julien Gagneur. «What we are doing with Kipoi is not just sharing data and software, but sharing models and algorithms that are already trained on the most relevant data. These models are ready to use, because all the cumbersome work of applying them to data has already been done,» says Anshul Kundaje, Assistant Professor at Stanford. More than 2,000 trained models are currently freely accessible on Kipoi. In a recent study published in Nature Biotechnology, the researchers show that the new repository will accelerate exchange in the genomics community and thereby advance genome research.

Fast algorithms and easy operation

Because Kipoi simplifies access to already trained models, researchers can perform transfer learning. This means that a model that has already been trained with a particular dataset is capable of learning a similar task faster. Kipoi also simplifies the process of feeding data into the models stored there: Standardized file formats and software frameworks reduce the installation and execution of a model to three simple commands. Those who previously had no experience in machine learning can thus also easily use the repository.

Understanding individual genomes

As Kipoi is oriented towards models that link genotype and phenotype, the new platform will make it easier to identify genetic causes of disease: «Kipoi puts the latest deep learning models trained on massive genomics data at the fingertips of clinical researchers,» says Julien Gagneur. «This provides very exciting opportunities to understand individual genomes, for instance to pinpoint genetic variants causing diseases or to interpret mutations occurring in tumors.»

However, the extent of the platform’s contribution to genomic research will also depend on the genomics community. «We hope that in the future more researchers will bring their models to our repository,» says Oliver Stegle, team leader at the EMBL-EBI. «That is the only way we can make genomics analysis accessible and ultimalely make a wider range of predictive machine learning tools available to the genomics community.»

Story Source: Materials provided by Technical University of Munich (TUM). Note: Content may be edited for style and length.