LFX - Researcher _ Proteomics quantitative analysis of type of sample

 
 

Samples

• 2 groups of x sample : Control (CTR) vs treatment (TRT)) (2 x 4 replicates) = 8 samples

• CTR_01
• CTR_02
• CTR_03
• CTR_04
• TRT_01
• TRT_02
• TRT_03

• TRT_04

   

Comparisons

• 4 CTR vs 4 TRT

Sample preparation and analysis

• 10ug proteins was extracted from sample
• Bradford assay was performed
• Proteins were reduced by DTT and alkylated with Iodoacetamide
• Proteins were digested with Trypsin (1:50) overnight at 37 degree celsius
• Resulting peptides were purified using StageTips C18 and Speed Vac dried
• Samples were resuspended in LC loading solvant
• 1 ug of each samples was analyzed by LC-MSMS on an Orbitrap Fusion Tribrid system (Thermo)
• 120 min LC runs (90 min grdients) were used and the MS operated in DDA mode
• Identification and LFQ quantification was done using MaxQuant software v1.6.10.43
• Database : Uniprot “Species” reference - Date_“x” entries
• Modifications: Fixed : Carbamidometyl (Cys), Variable : Oxidation(M), Acetyl (Prot N-term)

   

   

DATA TREATMENT (R software)

1 - Import & filtering data

• Input file : “proteinGroups.txt” from the MaxQuant software
• Decoy proteins and contaminant proteins are removed
• For data processing, only the LFQ intensity columns are considered

   

2 - Data normalization

• Normalisation is performed by the MaxQuant software; we use LFQ intensity.

   

3 - Missing data imputation

• For each sample, a noise value corresponding to the 0.01 percentile of the intensities of said sample is calculated
• This noise value is imputated for each samples when the intensity value is missing

   

4 - Comparisons

• Only proteins which present 100% of intensity values (not noise imputed value) in at least one group are considered. These proteins correspond to quantifiable proteins.
  
• Only proteins identified with at least 2 peptides are considered. They correspond to quantified proteins.

   

5 - Statistics

For each protein in each comparison, the following values are calculated :

• the mean intensity in each group (“mean_g1” and “mean_g2”),
  
• the intensity ratio between each group (“Ratio g1/g2”"),
  
• the log2 of g1/g2 ratio (“log2Ratio”")
  
• the fold change (= ratio si >1 ou 1/ratio si >1)
• the log2 fold change
• the z-score z = x – mean / standard deviation (“zscore”")
  
• Limma p-value and q-value (Benjamini Hochberg adjusted p-value) (“pval_Limma” and “qval_Limma”)
• The following filters were applied to define variant proteins : q-value < 0.05, |z-score| > 1.96 (“Sign_zscore_qval_Limma”)

   

       

DATA REPRESENTATION

1 - Check missing values

Low abundance proteins signal is usually not extracted in all samples. This graph shows the proportions of intensity values (grey) and missing values (white) for each sample of the analysis. We expect to observe low proportion of missing values (<15-20%) and a similar profile for all samples to compare.

• Between 13.83% and 20.75% of missing values between the samples
  
• 16.7% of missing values for the whole data set        

2 - Data normalization

Normalization was performed by MaxQuant software. LFQ intensities were used.

     

     

3 - PCA - Principal Component Analysis (proteins)

This representation uses an orthogonal transformation to convert all variables (here proteins) into new variables (principal component) explaining as much as possible the variability across the samples. The first component (PC1) always as the highest variance. Samples having similar proteomic profiles will appear close on the PCA plots. When the ellipses of two groups are not overlapping, we can consider that the sample groups can be distinguished based on their proteomic profiles.

     

4 - Proteomic profiles heatmap

This heatmap represents the intensity values in all samples of all proteins of the analysis. A hierachical clustering is applied both on rows and columns meaning that samples or proteins having similar profiles appear close to each other

5 - Number of quantifications for each comparison

Not all proteins of an analysis can be quantified, the signal of low abundance proteins might be difficult to extract resulting in missing values for certain sample. Proteins with a sufficient number of not missing values (100) are considered as quantifiable. Moreover, to add more confidence in the quantification value, we only retain proteins quantified with at least 2 peptides.

Identified proteins :
Total number of proteins identified in the analysis
506 proteins identified in this analysis

Quantifiable proteins :
Proteins with at least 100% of observed intensities in each replicates in one of the two groups.
412 quantifiable proteins on average for the whole analysis
 

Quantified proteins :
Proteins identified with at least 2 peptides.
380 quantified proteins on average for the whole analysis

     

6 - Number of variant proteins for each comparison

Proteins are considered as “variant” between two conditions/groups if they fulfill these criteria : q-value < 0.05 and |z-score| > 1.96 In a conmparison annotated “group1_group2”: the proteins “up-regulated” (red) are more abundant in “group1” than in “group2”, the proteins “down-regulated” (blue) are more abundant in “group2” than in “group1”,      

7 - Plots for each comparison

Repartition Histograms and Volcano plots

• The repartition histogram of all protein ratios (log2(ratio)), expected to be centered as much as possible to 0 (ratio g1/g2 = 1)
• The repartition histogram of z-scores (centering), must be centered to 0
  
• The volcano plot, where x-axis represents the protein ratio (log2(ratio)) and the y-axis the probability (-log10(p-value))
  The proteins are variant if they have a high ratio (in absolute value) and a good probability, i.e proteins in the upper right and upper left corners of the graph.

Box-plots and HeatMap of variant proteins

• The box-plots shows the proteins intensity value for all replicates of each group
• The heatmap shows the scaled intensity values for all variant proteins in all replicates, a hierachical clustering is applied both on rows and columns meaning that samples or proteins having similar profiles appear close to each other.