Improving Experimental Outcomes in Kinome Microarrays Through Quality Control

Abstract

Peptide microarrays consisting of defined phosphorylation target sites are an effective approach for highthroughput analysis of cellular kinase (kinome) activity. Kinome peptide arrays are highly customizable and do not require species-specific reagents to measure kinase activity, making them amenable for kinome analysis in any species. However, the data emerging from experiments with kinome peptide arrays exhibit a large amount of variability. To mitigate this issue, we introduce PIIKA 2.5 to expand upon existing software by providing three important quality control features in an aim to increase the accuracy and consistency of kinome results, which often suffer due to the aforementioned variability. The first feature concerns the size of the virtual circle drawn around each probe in microarray analysis software (spot size). This circle creates the boundary between pixels interpreted as foreground signal and pixels interpreted as background signal. In this thesis, it is shown that too large of a spot size creates abnormal data characteristics, such as high skewness (the asymmetry of the distribution of the data), that can alter downstream results. Here, a feature is presented that alerts users to the existence of improper spot size and informs them of the need to perform a manual alignment to enhance the quality of the raw intensity data, based on the skewness of the data as determined by examination of the mean and median of each dataset. The second feature uses interarray comparisons to identify outlier arrays that sometimes emerge as a consequence of technical or unknown issues. The work shown in this thesis indicates that the removal of said outlier arrays improves downstream analysis and interpretation. The third feature is a new background correction method, background scaling. Here, it is demonstrated to sharply reduce spatial biases in comparison to the most popular background correction method, background subtraction. Collectively, the modifications presented in PIIKA 2.5 allow users to identify low-quality data, improve clustering of treatment groups, reduce unintended effects, and enhance reproducibility in kinome analysis. The web-based and stand-alone versions of PIIKA 2.5 are freely accessible at http://saphire.usask.ca/saphire/piika.