Stepwise Distributed Open Innovation Contests for Software Development: Acceleration of Genome-Wide Association...

Area of Science:

• Bioinformatics
• Computational Biology
• Genetics

Background:

• Genotype-phenotype associations are crucial for identifying therapeutic targets and patient stratification.
• Advances in genotyping and phenotyping technologies have led to exponential growth in related datasets.
• Current genome-wide association study (GWAS) tools, like logistic regression, struggle to efficiently analyze these large datasets.

Purpose of the Study:

• To accelerate logistic regression analysis for large genotype-phenotype datasets.
• To leverage open innovation (OI) and contest-based crowdsourcing for software development.
• To improve the efficiency of genome-wide association studies (GWAS).

Main Methods:

• Employed open innovation (OI) and contest-based crowdsourcing to enhance existing genetics software (PLINK 1.07).
• Utilized a crowd-based contest to identify computational, numeric, and algorithmic optimizations.
• Integrated contest-derived code with parallelization and multithreading techniques for distributed innovation.

Main Results:

• Achieved an 18- to 45-fold acceleration of logistic regression within PLINK 1.07 through OI.
• Attained an overall end-to-end speedup of 591-fold by combining optimizations and distributed innovation.
• Reduced analysis time from 4.8 hours to 29 seconds for a dataset of 6678 subjects and 645,863 variants.

Conclusions:

• Developed a significantly faster logistic regression implementation for GWAS using iterative competition-based OI.
• Successfully applied OI to rapidly access specialized programming skills and accelerate bioinformatics tool development.
• Recommendations are provided for implementing successful OI processes in bioinformatics research.