Software project ’miraculix’: Efficient computations with large genomic datasets


  • Martin Schlather
  • Alexander Freudenberg University of Mannheim
  • Guido Moerkotte University of Mannheim
  • Torsten Pook University of Wageningen
  • Jeremie Vandenplas University of Wageningen


We present mathematical approaches for CPU accelerations to calculate matrix multiplications between a Single Nucleotide Polymorphism (SNP) matrix and another SNP matrix or a real-valued matrix. These accelerations are important in crucial time-relevant calculations of single-step evaluations and other methods in genetics. The presented algorithms are much faster than previous algorithms. The C-code is released as part of the software project ’miraculix’, which has been integrated into existing software such as MoBPS and MiXBLUP. We also discuss precision problems and missing SNP genotypes.