Changes between Version 3 and Version 4 of SOP/MassSpec

Timestamp:

12/06/24 11:28:30 (7 weeks ago)

Author:

gbell

Comment:

--

SOP/MassSpec

@@ -1 @@
-==== Differential protein expression (with mass spec) ====
+==== Preparing a mass spec experiment for differential protein expression ====
 
 This method is for label-free samples from our Proteomics Core Facility, which has some [[http://massspec.wi.mit.edu/documents/Scaffoldhowto060513.pdf|Scaffold quick instructions]].
@@ -41 +41 @@
     * To identify differentially expressed proteins, use the normalized Total Spectrum Count
       * Normalize across samples (typically using quantile normalization)
-      * Impute missing value (typically with the half-minimum method, which imputes any missing values for a protein with half of the minimum assayed value for that protein)
+      * Impute missing values 
       * Recommended statistic: t-test or moderated t-test (such as is implemented in 'limma') on log2 transformed values
       * Correct p-values with FDR (or an alternate method) 
@@ -51 +51 @@
       * Draw scatterplot using the normalized Top 3 Precursor Intensities, highlighting the differentially expressed proteins (from the Total Spectrum).  
 
+==== Preparing a mass spec experiment for differential protein expression ====
+
+More detail about this analysis: 
+
+  * Create a tab-delimited matrix of desired metric across all samples, with one column of unique protein identifiers
+
+  * Normalize by quantiles (or another method) across all samples, based on the assumption that total protein mass should be the same in each sample.  If this assumption is not valid, then spike-in (or another non-global) normalization method should be applied.  See our code: normalize_matrix.R (which also includes other methods).
+
+  * Impute missing values.  We prefer the half-minimum method, which imputes any missing values of a protein with half of the minimum assayed value for that protein.  This assumes that the true level of a protein with a missing value is between 0 and the minimum assayed level for that protein.  The half-min method calculates the middle of this range with our code: impute_missing_matrix_values.R (which also includes other methods).
+
+  * Calculate statistics for the differential expression analysis using limma, which applies moderated t-tests, one per protein.  The protein levels must first be log-transformed, but that step occurs within our code: Run_2_groups_limma_differential_expression.R (which also calculates adjusted p-values).  Choose an appropriate FDR threshold for differential expression.
+
+  * Create volcano and MA plots for a global perspective of changing protein levels.
 
 ==== Recommendations from Northeastern (May Institute, Vitek Lab)  ====