Gpu-enhanced scalable methods for genome sequence feature extraction and clustering

With the evolution of bioinformatics, vast amounts of genomic data are generated every day. Clustering has been widely used to derive meaningful insights from huge genomic datasets. To deal with such huge amounts of genomic data, scalable clustering algorithms were designed earlier. The main limitation of scalable clustering algorithms is that these methods cannot take the raw form of huge single nucleotide polymorphism (SNP) sequences as input. In this paper, we propose the Scalable GPU accelerated SNP feature extraction (SGPU-SNPfe) algorithm, which preprocesses the raw SNP sequences and produces twelve-dimensional numerical feature vectors. The SGPU-SNPfe enables Spark to utilize GPUs in high-performance computing (HPC). The preprocessed SNP sequences obtained from the SGPU-SNPfe algorithm are used as input to scalable fuzzy clustering algorithms. The experimental results demonstrate the effectiveness of the SGPU-SNPfe algorithm on scalable fuzzy clustering algorithms in terms of the Silhouette Index and Davies-Bouldin Index.