Parallelization and optimization of genetic analyses in isolation by distance web service

Area of Science:

• Population Genetics
• Bioinformatics
• Computational Biology

Background:

• The Isolation by Distance Web Service (IBDWS) facilitates population genetic data analysis using Mantel tests and reduced major axis regression (RMA).
• High computational demands of IBDWS lead to long processing times, hindering real-time analysis expectations.
• Increasing genetic data size necessitates improved processing power and efficiency.

Purpose of the Study:

• To enhance the speed and efficiency of the IBDWS by optimizing computationally intensive code sections.
• To investigate the potential for parallelization and algorithmic improvements within IBDWS.

Main Methods:

• Identified time-consuming components in IBDWS, specifically Mantel tests and RMA calculations.
• Restructured and parallelized code, including combining randomization routines and implementing the Fisher-Yates shuffling algorithm.
• Tested optimized algorithms on datasets ranging from 10 to 150 populations.

Main Results:

• Runtime tests revealed significant time consumption in Mantel test randomizations and RMA calculations.
• Optimized and parallelized routines demonstrated substantial runtime reductions.
• Achieved up to a 23-fold increase in analysis speed for datasets with 150 populations and 10,000 randomizations.

Conclusions:

• Implemented parallel algorithms have successfully improved IBDWS performance for large, real-world datasets.
• Continued growth in dataset size may require further hardware upgrades (e.g., cluster nodes).
• The parallelized IBDWS implementation is available at http://ibdws.sdsu.edu/.