Issues with Andromeda vs other search engines

[photocyte]

Hi there,

I am trying to replicate an analysis that some collaborators did with MaxQuant and (presumably, internally) Andromeda, using SearchGUI and PeptideShaker. This is because the GUI interface of SearchGUI & PeptideShaker allows me to more deeply understand the data.

Using SearchGUI, I can roughly replicate the results of the MaxQuant analyses (similar set of peptides detected, similar ions annotated within spectra), but only with X! Tandem and Comet. For Andromeda, there is a very little (essentially zero) overlap between it's identified peptides & the identifications between Comet and X! Tandem (which show a high degree of overlap)

To try to troubleshoot this, I toggled the Andromeda advanced parameters within SearchGUI, including the CID (default) vs HCD parameters, and the decoyMode parameter (None (default), or Reverse), over. 4 different runs. (For good measure, I also ran PeptideShaker with 1-core to avoid this known bug: https://github.com/compomics/peptide-shaker/issues/481 ). In all 4 runs, Andromeda behaved the same where it showed very little concordance with the X! Tandem and Comet results:

SearchGUI run | General settings | Andromeda - Fragmentation Method | Andromeda - Decoy Mode -- | -- | -- | -- v13 | 15 ppm precursor, 0.2 Da (Th) fragments | CID (default) | None (default) v14 | "" | HCD | None (default) v15 | "" | CID (default) | Reverse v16 | "" | HCD | Reverse

---

v13 - Screenshot of PeptideShaker Spectrum IDs tab. Note essentially all unique Andromeda PSMs vs X! Tandem, Comet

---

v14 - Screenshot of PeptideShaker Spectrum IDs tab. Note essentially all unique Andromeda PSMs vs X! Tandem, Comet

---

v15 - Screenshot of PeptideShaker Spectrum IDs tab. Note zero PSMs from Andromeda

---

v16 - Screenshot of PeptideShaker Spectrum IDs tab. Note zero PSMs from Andromeda

---

See this Google Drive link for the PeptideShaker .zip exports, for the 4 different runs: https://drive.google.com/drive/folders/1TuJcjHvsxyfrmxLA9CYu4EaCzLd0jOLx?usp=sharing

I'm using SearchGUI v4.2.17 and PeptideShaker v2.2.25 on Windows 10 Pro v. 22H2 . Java version (from command prompt) = openjdk version "11.0.16.1" 2022-08-12 LTS

My inputs to SearchGUI were centroided .mzML.

Is Andromeda via SearchGUI behaving correctly? Is there anything I can do to help troubleshoot?

All the best, -Tim

[hbarsnes]

Hi Tim,

I'm afraid that the Andromeda version included in SearchGUI is rather outdated as we have been waiting for years for an updated version of the standalone version of Andromeda. Hence I would not recommend spending too much time on optimizing the use of Andromeda, especially if you already have good results with some of the other search engines included in SearchGUI.

That being said, we haven't seen such poor results for Andromeda for other datasets as far as I can remember. So you could maybe to have a closer look at the other advanced Andromeda parameters to see if some of these can explain the very poor results for Andromeda. Maybe you can get the exact Andromeda parameters from the original search and see how well they match the ones used in SearchGUI?

Usually in such cases I would also recommend that you contact the search engine developers directly, but as the MaxQuant developers do not seem to have much interest in the standalone version of Andromeda there may not be much point to this.

Best regards, Harald

[photocyte]

Thank you Harald. In this case, since X! Tandem & Comet seem to work reliably, I will stick with those & simply remove Andromeda from my analysis.

You might consider a rating system in the SearchGUI interface or documentation, to highlight which search engines are the most mature. If Andromeda standalone has not had any interest from the MaxQuant developers, at some point presumably it must become deprecated from SearchGUI (or any other search engine where the upstream support drops off)