Reproducibility of cancer research