Targeted MRM versus discovery runs

Discovery runs name what is in the mixture. A targeted multiple-reaction-monitoring run, usually called MRM or SRM, measures peptides you have already decided to trust. The useful order is screen first and confirm second. Building a transition list for a protein nobody has shown to be present is possible, and it is a poor way to spend a calibration curve. The discovery logic is bottom-up proteomics in plain language. The digest that has to yield the watched peptide is preparing peptides for mass spectrometry.

Discovery names, targeted watches

A data-dependent or data-independent discovery method fragments peptides across a wide mass range and then asks a database which sequences fit. The result is a list with a false-discovery rate attached to the list. You learn candidates, including ones you did not expect, and you pay in missing values and in depth limits.

MRM on a triple quadrupole watches chosen pairs: a precursor mass and a fragment mass, called a transition. The instrument spends its time on that short list. Selected reaction monitoring is the same idea under the older name. Parallel reaction monitoring, PRM, is the related class on a high-resolution instrument: you still choose the precursors, and you record many fragments at once. PRM is not a discovery run. The peptide list existed before the queue started.

Because the list is closed, a decoy search is not the error model. A wrong fragment, a peptide that is not unique, or a matrix ion that happens to share the transition will look like a peak. Confidence is chromatographic and chemical: the traces co-elute, the fragment ratios match a standard, the retention time agrees, and a matrix blank is quiet.

Peptide, transitions, time and a standard

You need four things before the assay is honest. A stable peptide, meaning one the digest reliably makes and the tube does not ruin. A transition list. A retention time. And, if the sentence you want is quantitative in a calibrated sense, a calibration series or a labelled internal-standard class.

Choose a proteotypic peptide: fully tryptic if the digest is tryptic, unique in the database you will cite, and preferably free of methionine if oxidation would split the signal, or else monitored in both forms on purpose. Ragged ends and missed-cleavage variants scatter the signal. PeptideAtlas can show whether a peptide has been observed by others. UniProt is where you check that the sequence is unique to the accession you mean. Uniqueness in a different database version is not uniqueness in yours.

Transitions are instrument-specific. A triple quadrupole needs enough dwell time on each trace to describe the peak, and the peak must still be sampled across its width. Those times belong to the method file and to the manufacturer's guidance for that model. Do not copy a dwell table from a paper that watched four peptides into a method that watches forty. Retention time comes from a pure standard or from the discovery chromatogram on a matched column and gradient. A window that is too wide invites interference. A window that is too narrow misses the peak when the column ages.

Labelled internal standards are peptides with heavy amino acids, spiked before the step you need to control. They correct volume and ionisation for that sequence. They do not correct a peptide that was never produced because the digest failed. A calibration curve in buffer is weaker than a curve in the real matrix. NIST publishes reference-material thinking for measurement science. A research assay should say what your calibrant is traceable to in plain language, even when no formal reference material exists for that peptide. "We spiked a synthetic peptide of stated purity" is a clearer sentence than a vague standard.

A blank that fails, and what you do next

Run the matrix blank before the precious samples. The blank is the same biological background without the analyte, or the closest process blank you can make. If a transition lights up, retire it or change the fragment. Do not subtract a noisy blank and keep the transition for a low-abundance peptide.

If the standard in buffer works and the same standard spiked into matrix vanishes, the matrix is suppressing ionisation or shifting retention. Dilute, change cleanup, or move the gradient. protocols.io is a place laboratories record those cleanup variants. Follow the protocol you validated, not a neighbouring one.

If discovery saw the peptide and the targeted method does not, compare cysteine chemistry, enzyme and gradient length. A discovery alkylation that the targeted digest skipped changes the precursor mass. The transitions are then aimed at a peptide you did not make.

If all transitions co-elute except one, drop the discordant fragment. It is the interference. Keep at least the number of agreeing traces your assay plan required. An assay that quietly falls from five transitions to one is a different assay.

Interference and a wrong transition

The failure mode that looks most like success is a single smooth peak at roughly the expected time, produced by the matrix. The traces do not agree with each other, or they agree with the blank. Retire the transition. Adding more samples will not make the interference become the analyte.

A wrong peptide is the other quiet failure. The transitions are beautiful and the sequence is shared by two proteins, or it belongs to a contaminant that shares a stretch with your protein. Uniqueness is a database fact. Check it again when you change taxon or add isoforms. A heavy-labelled standard of the intended sequence co-elutes with the light peptide and supports identity. It does not prove the light signal is free of an overlapping matrix ion unless the fragment ratios still match.

Discovery false-discovery rates do not transfer. A peptide that passed a one-percent FDR in one search can still be a poor MRM target if it is short, shared, or sits in a dirty part of the gradient. Selection is a new experiment.

Research limits of a targeted number

An MRM peak is a research measurement under the assay you described. It is not a clinical concentration and not a diagnosis. If the matrix is human material, identifiers and ethics sit with your institution. Biosafety of any infectious lysate sits there too. Do not describe a research transition list as a cleared assay. Reference materials and formal measurement traceability, where they exist, are a different claim from a spiked synthetic peptide. Say which one you have.

Cold-chain handoff for a labelled peptide

Labelled peptides and dried digests are small amounts in a large story. A shipment that warmed in transit, or a vial that sat in a humid receiving room with a loose cap, can oxidise, adsorb or simply be wet. Record the temperature class on dispatch and on receipt, and run the standard before you run the cohort. If the standard's own transitions have collapsed, the samples cannot be calibrated against it. In a building where freezers share unstable power, a labelled stock needs a freezer history the same way a sample does. A curve from a degraded standard is a precise record of the degradation.

Whether the peptide list already exists

The enquiry that can be answered is specific. Do you already have peptides from a discovery analysis, with sequences and a database version? Do you need presence, a relative comparison, or a calibrated estimate? What is the matrix, and which labelled-standard class are you prepared to use? A list that does not yet exist means the next conversation is still a discovery design.

Use the shotgun discovery proteomics reference when the screen has not been done, the protein identification by LC-MS/MS reference when the need is still naming, and the differential abundance reference when the confirmation serves a contrast. Send the peptide list, or say plainly that it does not exist yet, with the quote request. The pages are references for that discussion. They do not mean a targeted queue is already running.