Table 1.

Existing Proteomics Imputation Methods^a

Method	Type	Description	Examples
Zero replacement	S	Replace missing values with zeros
Mean replacement	S	Replace missing values with the mean peptide intensity for a peptide or sample
Low value replacement	S	Replace missing values with the lowest observed intensity in any sample (sample minimum) or peptide (peptide minimum)
Gaussian random sample	S	Randomly sample from a Gaussian distribution centered around the lowest observed intensity	Perseus9
Regression	L	Linear regression is used to estimate missing values	lm, glm
kNN	L	Weighted average intensity of k most similar peptides	impute.knn,3 VIM10
MissForest	G	Nonparametric method to impute missing values using a random forest classifier trained on the observed parts of the data set, repeated until convergence	MissForest11
PCA	G	Run principal component analysis, impute missing values with the regularized reconstruction formulas and repeat until convergence	pcaMethods,12 missMDA13

^aDescriptions of generalized imputation strategies and examples of specific tools that implement each strategy. The “Type” column indicates whether the method uses single-value replacement (S), local similarity (L), or global similarity (G).