The batch effect is a big headache in high-throughput data (array or sequencing), especially when the data are generated in a long span of time. This includes detectable batch effect (e.g. by PCA, heatmap) and undetectable batch effect (noise).
But we can reduce the effect, for example, limit the running time within one week / one month, and maximize the running samples per day.
With the coming of Target capture + Ion Proton Sequencer + barcoding, we can measure multiple samples in each run for targeting genes (e.g. subtype predictor genes), and this surely minimize the batch effect and make the prediction more accurate.
