NEEP: null empirically estimated p-values for high-throughput molecular survival analysis

Null Empirically Estimated P-values (NEEP) is a non-parametric, high-throughput survival analysis meant for molecular expression data. The algorithm empirically finds the optimal cutoff (within a range) to separate expression values into low and high. The survival curves for this optimal cutoff are constructed. We calculate the MANTEL log-rank test for these optimally separated curves. The subsequent p-value distribution across all molecular objects will be skewed. To estimate the correct p-values, we sample 1M (or user specified) 'genes', which are a uniform permutation of the patient survival values from the provided clinical data. Then, p-values are estimated empirically and adjusted using the Benjamini-Hochberg approach.