median imputation or case deletion for small samples

I have a dilemma. I have a data set of different proteins from 4 groups of animals characterized by two factors: age (3 or 12 months) and presence of a certain gene (SPPL2B, KO or WT). We measured the 'abundance' of about 2000 different proteins in two brain regions of mice of different age and expressing or not the SPPL2B gene, which is important for the development of innate immunity. Because of not normal distributions we planned to apply ARTool to correct ranks for a nonparametric ANOVA. The nonparametric factor ANOVA was intended to verify the effect of age, that of SPPL2B, and possible interactions, on each of the 2000 proteins. For each combination of factors (3m, KO; 12m KO; 3m WT; 12m WT) I have 6 measurements (i.e., 6 animals), except for one group of which I have only 5 measurements. This prevents me from conducting an Aligned Rank Transform to then perform a nonparametric factorial (ANOVA) analysis. Instead, the analysis technically works if (a) I reduce the samples to 5 data points per group, or (b) if I replace the missing data point with the median of the 5 data points of the same group. I really don't know which one is the better solution.