Breeding programs from the construction kit

Henner Simianer, Lisa Büttgen, Amudha Ganesan, Torsten Pook
Animal Breeding and Genetics Group, Center for Integrated Breeding Research (CiBreed)

The primary task of breeding is to generate genetic progress, i.e. to modify livestock or crop populations in the long term so that they better meet the demands placed on them. This refers by far not only to higher performance, but also to improved health and resistance and higher quality of the products. In recent years in particular, adaptability to changing environmental conditions, better resource efficiency and lower emissions have also become important breeding objectives.

How can such diverse and even sometimes conflicting breeding goals be achieved? This is where breeding programs come in; these are complex processes planned and implemented by breeding associations or breeding companies. Typically, breeding programs consist of some central elements such as trait recording, selection of the best parents and their mating, supplemented by a multitude of additional process steps. Often, sophisticated statistical procedures and biotechnological methods are used in the context of complex logistics, whereby their concrete design always depends on the biological conditions in the respective plant or animal species. As a result, breeding programs today are as diverse as the species that are to be improved. For example, corn and potato breeding programs differ quite fundamentally, and the same applies to breeding programs in chicken or horse breeding.

At the Center for Integrated Breeding Research, we have been thinking about how to bring structure and order to this evolved diversity of breeding programs. Crucial to this was the realization that breeding programs are modular processes and can be represented as a directed sequence of only two types of modules, namely nodes and edges. Here, nodes are groups of individuals that have similar characteristics, e.g., a group of animals of the same age and sex for which similar performance information is available. Edges are processes that connect nodes, where the properties of the 'parent node' are transferred to the properties of the 'child node' according to certain rules. For example, a typical edge is selection, where only the best animals are selected from the parent node and transferred to the daughter node. Other edge types are e.g. 'aging' or 'reproduction'. The crucial insight now was that only an astonishingly small number of basic types of nodes and edges exist, which of course all have to be underpinned with further properties. These can be regarded as the elementary 'building blocks' for breeding programs. Any breeding program, no matter how complex, can then be represented in a modular way as a sequence of such nodes and edges, i.e. assembled from the various building blocks in a similar way to a Lego construction kit. In order to represent practical breeding programs, dozens or even hundreds of building blocks may be necessary, similar to the way many Lego blocks are needed to build, let's say, a realistic cow.

How can this insight now be used in practice? Based on the basic concept, we have developed the R package MoBPS (for Modular Breeding Program Simulator), with which breeding programs can be assembled according to the modular principle and then evaluated by means of stochastic simulation. This makes it possible to calculate the expected breeding progress in the various trait complexes or the expected development of inbreeding for a given breeding program. To enable easy use of the R package, we have developed a graphical user interface at www.mobps.de, which can be used to 'build' breeding programs from the individual modules in an intuitive way, and from which simulation can then be started via the R package at the click of a button. These programs have already been used in a variety of ways for scientific investigations into breeding planning in research projects, bachelor's and master's theses and doctoral projects, and are also used extensively in teaching and in cooperative projects with industry. The software was initially developed in an animal breeding context, but can in principle also be used in plant breeding. In a recently started PhD project, an extension with specific types of nodes and edges for plant breeding is planned. Furthermore, an 'optimization module' is to be developed, with which the breeding program structure and the use of resources can be optimized on the basis of MoBPS.

Interested parties are welcome to visit www.mobps.de for further information and to log in as a guest. There you will find various templates for breeding programs in the most important species, introductory documents and videos on how to use the software, and links to the scientific publications that have resulted from the project so far.

Original Publications: