How many spots with missing values can be tolerated in quantitative two-dimensional gel electrophoresis when applying univariate statistics?

Tuber proteome comparison of five potato varieties by principal component analysis

“…The occurrence of missing values may be due to numerous reasons, including: spot intensities below a threshold or limit of detection (LOD); mismatches caused by gel distortions; absence of spots as a result of experimental errors during transfer from first to second dimension; and very low abundance or actual absence of spots in the samples . The treatment of missing values in proteomic datasets is strongly recommended by several researchers . The main reason for this is primarily technical: most of the multivariate methods, either unsupervised or supervised classification tools, cannot deal with missing values …”

“…19,26 The treatment of missing values in proteomic datasets is strongly recommended by several researchers. 18,24,26,27 The main reason for this is primarily technical: most of the multivariate methods, either unsupervised or supervised classification tools, cannot deal with missing values. 20,28 To overcome this situation, a number of options for dealing with missing values have already been discussed.…”

“…The use of multivariate statistics such as PCA on proteomics data may be compromised by the occurrence of missing values in the dataset. There are no straightforward rules to deal with missing data, but elimination of missing values can be achieved by either excluding all variables containing missing values or by imputation of the missing data . Simply removing the variables would reduce the total number of spots considerably, given that 2‐DE data can have around 50% missing values .…”

“…There are no straightforward rules to deal with missing data, 26 but elimination of missing values can be achieved by either excluding all variables containing missing values or by imputation of the missing data. 27 Simply removing the variables would reduce the total number of spots considerably, given that 2-DE data can have around 50% missing values. 28 Imputation of data might be a better option, since specific algorithms can be applied for estimating quantitative values for missing spots based on the already existing spots.…”

A Practical and Analytical Comparative Study of Gel-Based Top-Down and Gel-Free Bottom-Up Proteomics Including Unbiased Proteoform Detection

“…A major challenge in proteomics analysis arises from missing detections of proteins, which reduce the number of comparable proteins in multiple analysis runs. Therefore, these missing values in the respective protein abundance data are one of the main problems in proteomics, as they severely impair the statistical evaluation of 2D gel and shotgun analyses and thus reduce the biological significance [49,50].…”