For a project, I want to predict the subcellular localization of some proteins. I am a student in software Informatics at the University of applied Science HEIG-VD.

Subcellular localization prediction is a common problem in the bioinformatics, but I have a problem

With the data I found. In this database: of human and mouse proteins, we can see that there are more than 30 different localizations.

At my level, and for the use of my exercise, 30 classes to predict is too much.

Where can I find a mapping to reduce those 30 classes to 4 or 5?

I tried QuickGO, to see the ancestor tree of the different classes, but nothing conclusive. Somebody might have done that, but I lack knowledge on bioinformatic web research.

There are the classes I wanted to map to reduce the number of classes :

  • Cytoplasm (GO:0005737)
  • Cytoplasmic Vesicles (GO:0016023)
  • Endoplasmic Reticulum (GO:0005783)
  • Endosomes (GO:0005768)
  • Extracellular (GO:0005576)
  • Golgi Apparatus (GO:0005794)
  • Lipid Particles (GO:0005811)
  • Lysosomes (GO:0005764)
  • Melanosome (GO:0042470)
  • Mitochondria (GO:0005739)
  • Nucleus (GO:0005634)
  • Peroxisome (GO:0005777)
  • Plasma Membrane (GO:0005886)
  • Synaptic Vesicles (GO:0008021)
  • Cellular Component Unknown (GO:0008372)
  • Apical Plasma Membrane (GO:0016324)
  • Basolateral Plasma Membrane (GO:0016323)
  • Centrosome (GO:0005813)
  • Golgi Cis Cisterna (GO:0000137)
  • Cytoskeleton (GO:0005856)
  • Early Endosomes (GO:0005769)
  • ERGIC (GO:0005793)
  • Inner Mitochondrial Membrane (GO:0005743)
  • Late Endosomes (GO:0005770)
  • Medial-Golgi (GO:0005797)
  • Outer Mitochondrial Membrane (GO:0005741)
  • Secretory Granule (GO:0030141)
  • Golgi Trans Cisterna (GO:0000137)
  • Golgi Trans Face (GO:0005802)
  • Tight Junction (GO:0005923)
  • Transport Vesicle (GO:0030133)

2 Answers 2


One way to do this would be to plot the first 10 GO categories of the the CC sub ontology (CC from cellular component). Or you can just visit the official webpage to see the list of children terms of them.

However there aren't only 4 or 5 categories, so you could arbitrary pick some of those.


This is probably one of those things where as long as you can justify your choice somewhat.

It might be worth seeing what others have done (see loctree3 and PSORT) and seeing how well they match your case, then add in any subdomains you have a particular interest in.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.