Andreas Gleiss1, Mohammed Dakna2, Harald Mischak2, Georg Heinze1. 1. Center for Medical Statistics, Informatics and Intelligent Systems, Medical University Vienna, Austria, Vienna, Austria and. 2. Mosaiques Diagnostics and Therapeutics AG, Hannover, Germany.
Abstract
MOTIVATION: A special characteristic of data from molecular biology is the frequent occurrence of zero intensity values which can arise either by true absence of a compound or by a signal that is below a technical limit of detection. RESULTS: While so-called two-part tests compare mixture distributions between groups, one-part tests treat the zero-inflated distributions as left-censored. The left-inflated mixture model combines these two approaches. Both types of distributional assumptions and combinations of both are considered in a simulation study to compare power and estimation of log fold change. We discuss issues of application using an example from peptidomics.The considered tests generally perform best in scenarios satisfying their respective distributional assumptions. In the absence of distributional assumptions, the two-part Wilcoxon test or the empirical likelihood ratio test is recommended. Assuming a log-normal subdistribution the left-inflated mixture model provides estimates for the proportions of the two considered types of zero intensities. AVAILABILITY: R code is available at http://cemsiis.meduniwien.ac.at/en/kb/science-research/software/
MOTIVATION: A special characteristic of data from molecular biology is the frequent occurrence of zero intensity values which can arise either by true absence of a compound or by a signal that is below a technical limit of detection. RESULTS: While so-called two-part tests compare mixture distributions between groups, one-part tests treat the zero-inflated distributions as left-censored. The left-inflated mixture model combines these two approaches. Both types of distributional assumptions and combinations of both are considered in a simulation study to compare power and estimation of log fold change. We discuss issues of application using an example from peptidomics.The considered tests generally perform best in scenarios satisfying their respective distributional assumptions. In the absence of distributional assumptions, the two-part Wilcoxon test or the empirical likelihood ratio test is recommended. Assuming a log-normal subdistribution the left-inflated mixture model provides estimates for the proportions of the two considered types of zero intensities. AVAILABILITY: R code is available at http://cemsiis.meduniwien.ac.at/en/kb/science-research/software/
Authors: Stuart R Lipsitz; Garrett M Fitzmaurice; Debajyoti Sinha; Alexander P Cole; Christian P Meyer; Quoc-Dien Trinh Journal: J R Stat Soc Ser C Appl Stat Date: 2020-02-03 Impact factor: 1.864
Authors: David L Tabb; Xia Wang; Steven A Carr; Karl R Clauser; Philipp Mertins; Matthew C Chambers; Jerry D Holman; Jing Wang; Bing Zhang; Lisa J Zimmerman; Xian Chen; Harsha P Gunawardena; Sherri R Davies; Matthew J C Ellis; Shunqiang Li; R Reid Townsend; Emily S Boja; Karen A Ketchum; Christopher R Kinsinger; Mehdi Mesri; Henry Rodriguez; Tao Liu; Sangtae Kim; Jason E McDermott; Samuel H Payne; Vladislav A Petyuk; Karin D Rodland; Richard D Smith; Feng Yang; Daniel W Chan; Bai Zhang; Hui Zhang; Zhen Zhang; Jian-Ying Zhou; Daniel C Liebler Journal: J Proteome Res Date: 2015-12-22 Impact factor: 4.466
Authors: Yuntong Li; Teresa W M Fan; Andrew N Lane; Woo-Young Kang; Susanne M Arnold; Arnold J Stromberg; Chi Wang; Li Chen Journal: BMC Bioinformatics Date: 2019-10-17 Impact factor: 3.169
Authors: Zhengyan Huang; Andrew N Lane; Teresa W-M Fan; Richard M Higashi; Heidi L Weiss; Xiangrong Yin; Chi Wang Journal: Sci Rep Date: 2020-01-21 Impact factor: 4.379