Quality control for a large-scale study using protein arrays and protein beads to measure immune response in serum...

We report on quality control performed in the context of a large-scale, multi-institutional study of the immune response in blood samples from prostate cancer patients. The measurements were performed by two commercially available techniques/services: protein arrays and an automated bead-based ELISA-like technique on 871 patient samples. The project started with a wide screen using standard arrays with 8302 protein sequences for 113 patients, followed by three studies using custom arrays with 215 selected protein sequences. These studies were followed up by three studies using the bead-based approach on 57 protein sequences chosen from the 215 selected before. We find similar responses in plasma and serum samples. Samples from the two European projects from which the samples originated also appeared comparable. In the data from the high-density standard arrays, we see for ~12% of the protein sequences high cross-correlation (R(2) > 0.8) with signals from unrelated protein sequences that are physically nearby on the array, suggesting production issues. The custom array and bead-based techniques both have good reproducibility, but the techniques do not agree with each other for ~50% of the protein sequences measured. We discuss the consequences of the observed data quality for the design and interpretation of the study.