Testing pulse detection algorithms with simulations of episodically pulsatile substrate, metabolite, or hormone release.

Mathematical computer-assisted simulations of episodic hormone, substrate, or metabolite release can be accomplished using explicit algebra and high-speed microprocessors. Such simulations are useful to describe the behavior of single or multiple pulse generators, as well as the expected output of these systems. Simulated series are helpful in evaluating the sensitivity, specificity, positive accuracy, and negative accuracy of discrete peak detection methods, and in deconvolution analysis. Specifically, deconvolution analysis should recover the correct half-life, production rate, frequency, amplitude, mass, and duration of pulsatile hormone secretion, and it should accurately estimate concurrent basal secretion, if present. Finally, multiple pulse generators provide an interesting simulation model for testing random burst concordance, evaluating how the surgelike release of a hormone can be accomplished, and assessing the interactions among several control systems. Lastly, special problems remain in simulating neuroendocrine pulsatility, namely, the impact of multiple binding proteins (see Table 1, p. 392 in this volume), the effects of strong and/or correlated circadian variations in burst frequency or amplitude or basal secretion, the development of improved statements of error and experimental uncertainty in the data, and the description of various modes of basal secretion.