A two-phase clustering procedure based on allele specific expression

Allele Specific Expression analysis is an important tool for integrating genome and transcriptome data. It quantifies expression variation between the two haplotypes of a diploid individual distinguished by heterozygous sites, and is a powerful tool to estimate cis-regulatory diversity of alleles. Clustering algorithms can be used to identify patterns or groups of genes/samples based on their expression profiles. Depending on the structure of the data, different existing clustering algorithm can be adapted to allele specific expression data. However, no ad-hoc procedure has been developed.

In this work, we begin defining an expression matrix capturing allele expressions from an RNA-sequencing experiment. On this matrix, we develop a novel two-phase unsupervised clustering procedure, built on top of a spectral clustering algorithm, whose aim is to partition the population into groups of similar individuals, according to their allelic expression. As case-studies, the approach is used to cluster 98 cultivars representative of the variability observed in Vitis vinifera, starting from read counts of genes of chromosome 1 of leaves, and to analyze allele-specific count data from a CASTxMRL F1 hybrid mice dataset.

Using the above mentioned real case-studies as well as generated synthetic data, we see that our algorithm shows significant robustness and outperforms other standard clustering techniques.