Table 1.

Experimental design and statistical considerations for maximizing the likelihood of reproducible biomarker discovery findings.

Study Design Step	Statistical Considerations
Study Goal	Define the population in which the protein marker will be used. Define the purpose of the protein marker—e.g., early detection, disease monitoring, etc.
Specimens	Select case and control specimens randomly. Select from a prospectively collected (prior to knowledge of disease status) specimen biobank. Select specimens from the relevant time point in the disease course. Avoid convenience samples. Avoid pooling of specimens.
Study Design	Plan sufficient sample size for discovery in light of realistic expected differences. Randomize specimens to assay run order.
Differential Abundance Detection	Assess protein difference signals relative to variation. Incorporate statistical design into the analysis model. Apply correction for multiple comparisons.
Panel or Signature Model Building	Finalize analysis plan in writing prior to beginning analyses. Employ optimism correction methods. Generate a fixed, locked-down algorithm.
Validation	Perform verification of initial protein marker identifications. Perform internal model validation in the discovery sample set. Perform external model validation in an independent sample set.