Local alignment tool based on Hadoop framework and GPU architecture.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Next-generation sequencing generates massive biological data, demanding high computational performance for analysis.
• Existing tools like SOAP utilize Hadoop and GPU architectures, but single GPUs are insufficient for large datasets.
• BLASTP is crucial for protein sequence comparison but faces performance bottlenecks with big data.

Purpose of the Study:

• To develop a distributed BLASTP system leveraging Hadoop and multi-GPU architectures.
• To address the computational challenges posed by large-scale biological data analysis.

Main Methods:

• Implementation of a distributed BLASTP system integrating Hadoop and multiple Graphics Processing Units (GPUs).
• Parallelization of the BLASTP algorithm across a cluster of GPUs managed by Hadoop.

Main Results:

• The distributed BLASTP significantly improves performance compared to single-GPU implementations.
• The system demonstrates high availability and fault tolerance, crucial for robust big data processing.
• Experimental results validate the effectiveness of the combined Hadoop and multi-GPU approach.

Conclusions:

• The proposed distributed BLASTP offers a scalable and reliable solution for analyzing large biological datasets.
• Combining Hadoop and multi-GPU computing is an effective strategy for accelerating bioinformatics tasks.
• This approach enhances the efficiency and resilience of protein sequence comparison in the era of big data.