Spell-QTL, a New Tool for QTL Analysis on Modern Datasets

Spell-QTL (https://mulcyber.toulouse.inra.fr/projects/spel/) performs classical QTL analysis in modern datasets. Complex pedigrees have recently emerged, such as MAGIC8, that make classic population models obsolete. Inferring Parental Origins has become a tedious task and genotype data is often raw SNP data. We implemented innovative methods to handle virtually any pedigree and precisely compute the Parental Origin Probabilities (POP) at any locus given any number of observed markers.

We construct a Bayesian Network matching the structure of the pedigree. The exact POP at each observed marker are inferred using this network. Typically, marker observations involve the ancestors and the phenotyped generation, but our method can take observations on any individuals and produce POP for any subset of individuals. Along a linkage group, the probabilities of recombination in each generation are modelled with Continuous-Time Markov Chain (time here being genetic distance) which enable precise computation of the POP at any locus, taking into account all the information provided by the neighboring markers. The QTL analysis is performed with a classical model selection approach.

The software suite is divided in three distinct binaries. Spell-pedigree analyzes the pedigree and computes the CTMCs. Spell-marker computes the POP at the markers. Spell-QTL performs the QTL detection and analysis. We emphasize speed and allow concurrency or even distributed computations wherever appropriate. We provide a rich and fully-documented command-line interface and the results are output in text files for convenient import in R or Excel.