Allegro, a new computer program for multipoint linkage analysis

• Published in: Nature genetics Vol. 25; no. 1; pp. 12 - 13
• Main Authors: Gudbjartsson, Daniel F, Jonasson, Kristjan, Frigge, Michael L, Kong, Augustine
• Format: Journal Article
• Language: English
• Published: United States Nature Publishing Group 01.05.2000
• Subjects: Algorithms; Analysis; DNA sequencers; Female; Genetic Linkage; Genetic Markers; Humans; Innovations; Linkage (Genetics); Lod Score; Male; Pedigree; Software - statistics & numerical data
• ISSN: 1061-4036; 1546-1718
• DOI: 10.1038/75514

The first generally available computer program for linkage analysis was Liped, introduced in 1974. It calculates two-point parametric lod scores for general pedigrees using the Elston-Stewart algorithm. Later programs, including Linkage, Fastlink and Vitesse, can calculate multipoint lod scores for a few markers, but execution time increases rapidly with the number of markers. Exact multipoint linkage calculations involving many markers for general families was first made practical in 1996 with the introduction of the program Genehunter. The run time of Genehunter grows linearly with the number of markers, but exponentially with pedigree size (p1 in Fig. 1a is close to the largest pedigree that can be analysed). In addition to parametric analysis, Genehunter also implemented non-parametric linkage (NPL) scores. These, however, can be conservative when sharing information is incomplete; to adjust for this, allele-sharing lod scores based on one-degree-of-freedom models were introduced and implemented in a modified version of Genehunter, Genehunter-Plus. We have developed several improvements to the computational algorithms of Genehunter. The new algorithms have been incorporated in a computer program for multipoint linkage analysis, Allegro, which is available free for non-commercial use (e-mail: allegro sub(e)code.is). Included is a program manual, a technical report and the source code. Like Genehunter, Allegro uses a hidden Markov model, and time and memory costs still grow exponentially with pedigree size, but it is considerably faster than Genehunter (typically 20-100 times, Table 1). Apart from allowing for larger pedigrees (typically 20-30% larger), the speed improvement is relevant for simulation studies. Instead of providing the technical details of the computational algorithms, here we describe Allegro from the user's perspective.