A glossary of proteomics terms

The words on a proteomics table decide whether a pull-down is defendable. The same words decide whether a treated-versus-control study can carry a sentence about abundance. Loose synonyms turn a shared peptide into a specific isoform, or a processing day into a pathway. This glossary walks the terms in the order a result is built, using those two applications as the test. The surrounding method is bottom-up proteomics in plain language. A pull-down only means something if the enrichment itself was real; affinity and ion exchange protein purification is the upstream language for that separation.

From a precursor ion to a fragment pattern

A precursor is the intact peptide ion the instrument measures before it breaks the peptide apart. Its mass-to-charge ratio and its charge are the address of that peptide in the run. In a pull-down, an abundant precursor from the antibody or the bait can occupy the instrument so thoroughly that partners are rarely chosen for fragmentation. If you claim a partner was absent, say whether its precursor mass was even in the method's range and whether the bait dominated the precursors. Absence of a name is not absence of a protein when the precursor list was full of something else.

A fragment is a piece of that peptide after fragmentation. Collision-based methods often produce series that read along the sequence. The fragment pattern is what a search engine compares with a theoretical pattern. A precursor mass that happens to be close to a calculated peptide is not identification. The fragments have to agree. When two peptides of the bait's family share a precursor mass, the fragments are the only reason to prefer one sequence. If the spectrum is chimeric, which is common when a pull-down is simple but still co-elutes, say so rather than forcing a single clean story.

A peptide-spectrum match, a PSM, is one spectrum paired with one sequence at a score. Counts of PSMs are evidence of how often that sequence was sampled, not a calibrated copy number. In a differential study, a protein with three PSMs in one sample and thirty in another is a hint, and it is still not a fold change until the quantification method says how intensity, not merely match counts, was compared. Report PSM counts as sampling. Report abundances from the method you actually used.

Assigning peptides to protein groups

A protein group is the set of database accessions that explain the same observed peptides. Search engines build groups because isoforms, splice forms and recent duplications share sequence. In the pull-down example, "the partner" may be a group of three accessions. The honest sentence names the group. UniProt and the NCBI Protein database will show you how those accessions are related. They will not invent a unique peptide you did not observe.

A unique peptide is a sequence that matches only one protein in the database you searched. It is the peptide that lets you point at one accession. Uniqueness is not a property of nature alone. A smaller database makes more peptides look unique. A database full of isoforms makes uniqueness rare. State the database when you call a peptide unique. In the differential study, a treatment effect assigned to one isoform needs a unique peptide that itself carries the quantitative signal, or you are reporting the group.

A razor peptide is a shared peptide assigned to the protein group that already has more independent evidence, an application of parsimony. It keeps the peptide's intensity from being counted twice, and it can place abundance on a group the peptide does not uniquely prove. In the pull-down, a razor peptide may inflate the apparent evidence for the bait's favourite paralogue. Read the column that says whether the peptide is unique or razor before you write the isoform into the paper. If the software does not provide that column, ask for a rerun that does. Do not guess.

Thresholds, sites and the mixture you cannot see

The false discovery rate, FDR, is an estimate of the fraction of accepted matches that are wrong, judged by searching decoy sequences alongside the real database. It applies to the list at a chosen score cut. A one-percent PSM FDR does not mean each row is ninety-nine percent certain in ordinary speech, and a protein-level FDR is a second cut on a different list. In the pull-down, a single-PSM partner that sits near the cut is the first name you should be willing to lose. The Proteomics Standards Initiative is where reporting expectations for these thresholds are discussed. The Human Proteome Organization is where the community argues about protein-level evidence. Neither body will choose your cut-off for you.

Localisation probability is the chance, given the fragment ions, that a modification sits on one residue rather than another. It matters when the pull-down is being used to claim a phosphosite on the partner. A peptide can be truly phosphorylated and still have an ambiguous site if two serines are close and the fragments do not bracket them. Report the site only when the localisation figure meets the rule you set before looking at the hits. Otherwise report the modified peptide without a residue number.

Dynamic range is the span of abundances the method can see in one injection. The bait, the antibody and serum albumin from a wash that was not harsh enough sit at the top. The partner you care about may sit below the point where precursors are chosen. A differential cell study has the same shape: a few abundant metabolic proteins and a long tail of scarce regulators. "Not detected" inside a wide dynamic range is a visibility statement. It becomes a biological absence only with a method designed to see that protein, such as a targeted assay.

A batch effect is a technical signature shared by samples that were processed or injected together. In the pull-down series, all the treated lysates digested on Monday and all the controls on Tuesday will separate by day. Any partner that is higher in the treated group is then also higher on Monday. Call it a batch until the design says otherwise. A plot that colours points by processing day is the glossary term made visible.

A contaminant is a protein that entered from skin, dust, the enzyme, the medium or the affinity reagent rather than from the biology you intended. Keratin, trypsin, streptavidin and antibody chains are the usual cast in a pull-down. They are still real PSMs. Flag them with a contaminant collection in the search. Deleting them silently changes total-intensity normalisation in the differential study, because you have removed signal that was part of the scaling. Flag, do not quietly drop, unless the report states the rule.

Two claims the glossary will not support

The pull-down claim "isoform 2 is the binder and it is phosphorylated on serine 114" needs a unique peptide for isoform 2, fragments that localise the phosphate, a protein group that is not only the antibody, and a control enrichment that lacks the partner. Missing any one of those, the strongest sentence is that a peptide shared by the family was modified somewhere in that stretch and was enriched with the bait under this protocol.

The differential claim "the treatment activates the pathway" needs a designed contrast, a named statistical comparison, and a check that batch and contaminants do not create the separation. A cluster of protein groups moving together is a hypothesis list. Pathway membership was not measured by the mass spectrometer.

Research use of the vocabulary

Use these terms in laboratory records and in research papers. Do not use them to imply a diagnosis or a mechanism the design did not test. Biosafety of the cells or tissues you lysed remains an institutional decision. A careful protein group does not change the hazard of the sample, and a contaminant flag is not a disinfection step.

Humidity, dust and the word contaminant

On a humid bench, keratin and dust stick to tubes and to gloves that touched a face shield. Call those proteins contaminants in the table even if the room feels clean. The decision the word changes is practical: a partner protein at the same intensity as keratin in a single pull-down is not yet a partner. Wipe-down and covered tubes are handling. The glossary still requires the flag in the search so the next person sees the same names.

Asking for a table that uses these words

When you ask for an analysis, ask for columns that distinguish unique and razor peptides, for the FDR cuts, for localisation scores where a site is claimed, and for a contaminant flag. Ask for the database version, because uniqueness depends on it.

Those requests sit naturally beside the shotgun discovery proteomics reference, the protein identification by LC-MS/MS reference and the differential abundance reference. Point at the one that matches the application. Send the bait system or the contrast, and the words you need in the table, through the quote request. The pages are a way to discuss the deliverable. They do not mean a pull-down has already been acquired.