Proteomics quantitative analysis in Sacsin depleted cell model
To replicate the differential abundance analysis done in the manuscript, run the "lfq_analysis_limma_pub.R" script file with R, using the data in "proteomicstable.txt" (both files must be in the working directory of the R session).
Methods description:
LFQ values were used for protein quantification. Only proteins with at least two LFQ values in at least one of the replicate groups (KO or WT) were kept for further analysis. Additionally, only proteins with more than one unique peptide detected were kept. LFQ values were log2 transformed and normalized by the median of the subgroup of proteins that were detected in the six samples. Two methods of missing value imputation were applied. When the protein was detected in two replicates, the missing value in that replicate group was defined with the k-nearest neighbors (knn) algorithm, using the average values of the five proteins (k=5) more similar (euclidean distance between LFQ values across samples) to the protein with the missing value. When the protein was detected in less than two samples in the replicate group, the missing values were defined as random values from a normal distribution with mean equal to the minimum LFQ of that sample and standard deviation equal to a global estimate made with all proteins with replicate values. It was previously checked that standard deviation estimates were not significantly affected by protein average abundance or number of detected peptides. Statistical testing of differential abundance between the two conditions was performed with the limma R package (Ritchie et al., 2015). Adjusted p-values lower than 0.05 were considered significant (False Discovery Rate (FDR) - 5%). To avoid artifacts induced by the missing value imputation with random numbers, this imputation was repeated 30 times, and only proteins with differential abundances consistently (>50% of the times) considered statistically significant were selected for further analysis.