Abacus

Instructions for running Abacus on UWPR systems

Abacus is a TPP-compatible spectral counting tool developed by the Nesvizhskii lab at the University of Michigan. Here are instructions for running it on the command line on our Linux systems.

Running Abacus

Generate individual interact-*.pep.xml and interact-*.prot.xml files for each of your input groups.

For a separate output column per search run, the easiest way is:
```
runCometQ --wocomet --single *.mzXML
```
or, if you are using decoys in the search:
```
runCometQ --wocomet --single --decoy DECOY_ *.mzXML
```
Otherwise, run a command like the following for each of your files:
```
xinteract -OAp -Nfile1 file1.pep.xml
xinteract -OAp -Nfile2 file2.pep.xml
xinteract -OAp -Nfile3 file3.pep.xml
xinteract -OAp -Nfile4 file4.pep.xml
```
To group runs so that each spectral count column covers two or more runs, generate the interact files for the groups, giving the groups meaningful names (e.g. “group1” and “group2”):
```
xinteract -OAp -Ngroup1 file1.pep.xml file2.pep.xml
xinteract -OAp -Ngroup2 file3.pep.xml file4.pep.xml
```
Generate a combined ProteinProphet file using all of the interact-*.pep.xml files as input:
```
ProteinProphet interact*.pep.xml interact-COMBINED.prot.xml
```
You should now have a set of interact-*.pep.xml and interact-*.prot.xml files plus interact-COMBINED.prot.xml. Abacus attempts to load every file in the directory, including the base search files (e.g. file1.pep.xml above). Temporarily move those into a “bak” subdirectory; this keeps them out of the analysis and shortens the run time. Leave the interact-*.pep.xml files where they are.
```
mkdir bak
mv file1.pep.xml file2.pep.xml file3.pep.xml file4.pep.xml bak/
```
Grab a copy of the Abacus parameters file:
```
cp /net/pr/vol3/www/html/protocols06/Abacus/Abacus_parameters.txt .
```
This parameters file is set to run Abacus with default settings using NSAF. If you need help understanding the settings, see the Abacus support forum. You will need to edit these entries:
- combinedFile: the full path to the interact-COMBINED.prot.xml file from step 2
- srcDir: the directory that contains your files
- outputFile: where you want the Abacus output to go
- fasta: the fasta file
- decoyTag: currently set to “DECOY_”; change as needed

Run Abacus:

java -Xmx32g -jar /net/pr/vol1/ProteomicsResource/bin/abacus.jar -p Abacus_parameters.txt

Undo step 3:
```
mv bak/* .
rmdir bak
```
That’s it. You should now have an ABACUS_output.tsv (or whatever you named the output file in step 4) in your search directory.

ADJNSAF

Abacus exports ADJNSAF (adjusted normalized spectral abundance factor), a label-free, spectral-count-based metric used in shotgun proteomics to estimate relative (not absolute) protein abundance. It extends the normalized spectral abundance factor (NSAF) with adjustments that improve comparability across samples, particularly in affinity-purification mass spectrometry (AP-MS) and interactomics experiments. The description below is ChatGPT’s.

Why it is needed. Raw spectral counts are biased: longer proteins generate more tryptic peptides and therefore more spectra, the total number of MS/MS spectra differs between runs, and highly abundant proteins can dominate the signal and obscure lower-abundance proteins. NSAF corrects for protein length and total spectral depth. ADJNSAF further corrects for run-to-run compositional effects and background dominance.

The NSAF calculation. For each protein i, count the number of MS/MS spectra assigned to that protein, typically including unique and shared peptides and filtered at a fixed false discovery rate (e.g. 1% FDR):

SC_i = number of spectra matched to protein i

Correct for the protein length L_i (in amino acids):

SAF_i = SC_i / L_i

Normalize across all proteins in the run, so that the values sum to 1 and proteins can be compared within the same sample:

NSAF_i = SAF_i / Σ SAF_j

The adjustment step. Abacus improves spectral count quantitation by explicitly accounting for shared peptides. Instead of assigning the full spectral count of a shared peptide to every protein it maps to, Abacus distributes those spectra across proteins based on the amount of unique spectral evidence supporting each protein:

Count the unique spectral evidence for each protein i:

s_i = number of spectra from peptides unique to protein i
For a peptide p shared among N proteins, compute an adjustment factor representing the proportion of the shared peptide’s spectral evidence attributed to protein i. Proteins with little or no unique evidence receive little or no contribution:

a_p,i = s_i / Σ s_j (j = 1…N)
Sum the unique spectra and the weighted contributions from shared peptides:

AdjustedSC_i = Σ unique spectra + Σ (shared spectra × a_p,i)
Compute NSAF using the adjusted spectral counts:

NSAF_i = (AdjustedSC_i / L_i) / Σ (AdjustedSC_j / L_j)

When NSAF is calculated using adjusted spectral counts, it is commonly referred to as ADJNSAF. This produces a more realistic estimate of protein abundance, especially for homologous proteins and protein families.

Interpretation. Higher ADJNSAF indicates higher relative protein abundance. Values are meaningful for comparisons within and across samples, but do not represent absolute protein concentration.

Strengths. No isotope labeling required, robust to missing MS1 intensity values, and well suited for interactomics and large protein complexes.

Limitations. Lower dynamic range than MS1 intensity-based methods, reduced accuracy for very low spectral counts, and dependent on digestion efficiency and peptide detectability.

Common applications. Affinity-purification mass spectrometry (AP-MS), protein-protein interaction studies, and semi-quantitative analysis of large proteomics datasets.