Table 2.
Abbreviated names of the subsequent processing methods used in this study together with their corresponding full names and brief descriptions
| Processing method (abbreviation) | Brief description of each method | |
|---|---|---|
| Transformation | Box–Cox Transformation (BOX) | Transforming data based on linearity, normality and homoscedasticity assumption [76] |
| Cube Root Transformation (CUB) | Improving normality distribution of simple count data [77] | |
| Log Transformation (LOG) | Almost routinely carried out for reaching a more symmetricdistribution [78] | |
| None (NON) | No transformation method applied | |
| Power Transformation (POW) | Stabilizing variance and making data more normal distribution-like [79] | |
| Normalization | Auto Scaling (ATO) | Scaling protein intensities based on the standard deviation of OMICdata [80] |
| Cyclic Loess (CYC) | Estimating a regression surface using multivariate smoothingprocedure [81] | |
| EigenMS (EIG) | Preserving true differences based on treatment effect and singularvalue decomposition [50] | |
| Linear Baseline Scaling (LIN) | Mapping each spectrum to the baseline based on a constant linearrelationship [82] | |
| Locally Weighted Scatterplot Smoothing (LOW) | Normalizing a proteomic data set by compensating for non-linearbias [83] | |
| Mean Normalization (MEA) | Normalizing the data by mean value of all signals to eliminatebackground effect [84] | |
| Median Absolute Deviation (MAD) | A measure of the spread of the data and used to estimate the samplestandard deviation [85] | |
| Median Normalization (MED) | Scaling the samples so that they have the same median [86] | |
| None (NON) | No normalization method applied | |
| Pareto Scaling (PAR) | Reducing the weight of large fold changes in protein intensities bystandard deviation [87] | |
| Probabilistic Quotient Normalization (PQN) | Transforming the spectra based on an overall estimation on the mostprobable dilution [88] | |
| Quantile Normalization (QUA) | Achieving the same distribution of protein intensities across allsamples [82] | |
| Robust Linear Regression (RLR) | Used for transference to rescale one reference interval to anotherscale [50] | |
| Total Ion Current (TIC) | Summing all the separate ion currents carried by the ions of differentm/z [89] | |
| Trimmed Mean of M Values (TMM) | Estimating scale factors between samples for differential expressionanalysis [90] | |
| Variance Stabilization Normalization (VSN) | A non-linear method for keeping the variance constant over the entiredata range [50] | |
| Z-score normalization (ZSC) | Normalizing data based on the mean and standard deviation [92] | |
| Imputation | Background Imputation (BAK) | Simulating the situation where protein values are missing [44] |
| Bayesian Principal Component Imputation (BPC) | Providing capacity to auto-select the parameters used in theestimation [44] | |
| Censored Imputation (CEN) | Imputing non-missing completely at random missing values by lowestintensity value [44] | |
| K-nearest Neighbor Imputation (KNN) | Imputing values based on K proteins similar to the proteins withmissing values [93] | |
| Local Least Squares Imputation (LLS) | Missing value imputation by a linear combination of similar genesidentified by KNN [44] | |
| None (NON) | No imputation/filtering method applied | |
| Singular Value Decomposition (SVD) | Estimating the missing values based on a linear consideration [93] | |
| Zero Imputation (ZER) | Replacing the missing values with zeros deemed to the simplestimputation method [94] |