Genomic Prediction

Machine learning-based breeding values.


Augment your phenotype-based program by automatically screening for resistances, qualitative, and quantitative traits. This enables you to make well-informed decisions about selection and deselection.
Derive value from the results we obtain: Insightful, interpretable visualizations.

Benefit from our machine learning-based regularized kernel methods to predict phenotypes from genome-wide markers. These methods model heterosis and genetic gain. We store the trained predictors to reproducibly analyze next season’s data and make results directly comparable.


  • Predict performance: Learn how your cross will perform before it is field-tested.
  • High throughput: Genotype millions of genetic markers in thousands of plant lines.
  • Fast improvement: Receive machine learning-based breeding values and phenotypes within 48 hours and improve the model with each cycle.

Client project: Reference-free genotyping of four stages

We performed reference-free genotyping of four stages of a breeding program to determine which locations are most predictive for the entire region. This reduced the number of regions needed for early- stage selection from 11 to just 3, freeing up space and saving costs.

More Information


Tübingen, Germany

Computomics is based in the university town of Tübingen, situated in the Southwest of Germany, but serves clients all over the world. We also have offices in Davis, California and Madison, Wisconsin.

Please calculate 7 plus 8.

This form collects your name, e-mail address and message so that we can get in touch with you about your inquiry. Please see our privacy policy below about how we protect and manage your submitted data.


Computomics GmbH
Christophstr. 32
72072 Tübingen

Phone: +49 7071 568 3995