Research question	High-dimensional problems
Biomarker selection (find individual genes or groups of genes that are functionally relevant and/or surrogates for a specific clinical outcome).	Trade-off between accuracy and computational complexity; confound of multimodality; spurious correlations; multiple testing; insufficient sensitivity of criterion function; curse of dimensionality; model overfitting.
Cancer classification (find new molecular subclasses within a cancer that exhibit meaningful clinical or biological properties, for example, that enable physicians to better direct treatment).	Curse of dimensionality; confound of multimodality; spurious clusters; model overfitting; small sample size; biased performance estimate.
Cancer prognosis (predict which patients will have a particularly good or poor outcome and enable physicians to better direct treatment).	Curse of dimensionality; confound of multimodality; spurious correlations; model overfitting; small sample size; biased performance estimate.
Cell signaling (identify how cell signalling affects cancer cell functions and identify new targets for drug development).	Curse of dimensionality; confound of multimodality; spurious correlations; multiple testing.
Predicting drug responsiveness (identify genes or patterns of genes that will predict how a cancer will respond to a specific therapeutic strategy and enable physicians to better direct treatment).	Curse of dimensionality; confound of multimodality; spurious correlations; model overfitting; small sample size.