Targeted proteomics with LC-MS/MS — from SRM to biomarker report

Targeted proteomics by LC-MS/MS is the quantitative gold standard for protein biomarkers, replacing ELISA where multiplexing, dynamic range, and freedom from antibody cross-reactivity matter. The measurement principle is elegant: tryptic peptides are fragmented, and the ratio of analyte to stable-isotope-labeled (SIL) internal standard is read as a surrogate for protein concentration. The challenge is that every step from sample prep to peak integration has a failure mode that can silently bias the result. This piece walks through the pipeline that makes the numbers auditable.

Assay design — where most time is lost

Peptide selection

The choice of surrogate peptide determines almost everything: ionization efficiency, recovery, interference susceptibility, and cross-reactivity to related proteins. Select peptides that are unique (BLAST against the proteome), free of missed cleavages, Met-free (or oxidation-monitored), deamidation-free at Asn-Gly and Asn-Ser, and that produce abundant b- and y-ions in the 350–1,200 Da precursor window. Screen at least five candidates per protein and carry forward only those with a CV < 15% across three matrix lots.

Stable-isotope internal standards

SIL-peptide (ⁱ³C/¹⁵N on C-terminal Lys or Arg, minimum +6 Da shift) remains the most practical IS for regulated assays. Proteins-labeled at isotopic full enrichment (stable-isotope-labeled protein standard, SILAC-based) give superior recovery tracking but cost more and require a spike-in at the earliest sample-prep step — mandatory for assays that need to correct for protein-extraction variability.

The sample-prep reproducibility problem

Reduction, alkylation, and trypsin digestion introduce three large sources of variance: incomplete reduction, variable digestion efficiency, and matrix-dependent recovery. A digestion efficiency > 95% at each step is the target; check it with a surrogate peptide that flanks the target peptide and would appear only if the preceding cleavage site was missed.

• Reduction: TCEP or DTT at 56 °C, 60 min; verify with a cysteic acid peptide.
• Alkylation: iodoacetamide in the dark; excess quenched before trypsin addition.
• Digestion: trypsin enzyme-to-substrate ratio 1:25 to 1:50; 37 °C, 16 h; inhibit with formic acid to 1%.
• Enrichment: SPE or SCX cleanup per matrix — plasma needs phospholipid removal (protein precipitation or dedicated cartridge) or signal suppression compounds will dominate.

Calibration and quantification

The FDA Bioanalytical Method Validation Guidance (2018) and EMA equivalent are the reference documents for quantitative proteomics in biopharma contexts. ISO 15189 applies in clinical labs. Both require:

• Calibration curve: ≥ 6 non-zero levels spanning LLOQ to ULOQ, prepared in surrogate matrix or stripped matrix matched to clinical samples. Fit a weighted linear or quadratic model (1/x or 1/x² weighting typical). Back-calculated concentrations within ± 15% (± 20% at LLOQ).
• LLOQ: lowest level with accuracy 80–120% and CV ≤ 20%, verified across ≥ 5 runs.
• ULOQ: highest level on the linear portion; samples above must be diluted.
• Quality controls: low, mid, high QC in ≥ 5 replicate per run; 4/6 within ± 15% and no more than 1/6 outside at any level.
• Matrix effects: post-column infusion or matrix-factor experiment across ≥ 6 matrix lots from different donors, including lipemic and hemolyzed where clinically relevant.
• Dilutional linearity: high-concentration sample diluted 2×, 4×, 8× — accuracy must hold after application of the dilution factor.
• Stability: freeze-thaw (3 cycles), long-term frozen, bench-top, post-preparation, autosampler.

Peak integration — the hidden operator variable

Manual peak-integration review is the largest remaining source of between-analyst variability in targeted proteomics. Automated integration with a peak-review SOP is the only scalable solution:

1. Lock integration parameters (peak-find algorithm, baseline, smoothing, minimum signal-to-noise) in the SOP before the analytical run.
2. Flag peaks with signal-to-noise < 9, asymmetry factor > 2.0, or co-eluting interference.
3. Record every manual override with operator ID, timestamp, and reason — these overrides are the first thing an auditor reviews.

Multiplexed biomarker panels

One of the primary advantages of LC-MS/MS over ELISA is the ability to quantify 10–50 proteins in a single injection. Multiplexed panels require additional QC discipline:

• Inter-peptide carry-over: a high-concentration peptide from one protein can contaminate the next injection if not washed correctly. Run a blank after every high calibrator.
• Co-elution conflicts: two target peptides with overlapping retention and a shared fragment ion produce false addition. Pre-screen for conflicts in silico before instrument time is allocated.
• Relative ionization suppression: one peptide in the panel can suppress another in complex matrices. Verify by comparing apparent concentrations in neat buffer vs. spiked matrix for each peptide.

Where labs lose regulatory approvals

• Surrogate matrix mismatch: validated in BSA buffer, deployed in clinical plasma — the matrix effect was never characterized.
• Unlocked software versions: peak areas computed differently after a software update, but the SOP was not revised.
• Missing IS response tracking: the IS peak area is the single most sensitive indicator of systematic error. If it is not trended over every run, drift goes undetected until QC samples fail.

How AiLabrix fits

Drop the instrument export (Skyline, Xcalibur, LabSolutions) plus the sample metadata. The pipeline parses peak areas and IS responses, applies weighted linear calibration, computes LLOQ/ULOQ, runs the FDA BV acceptance criteria per analyte, generates matrix-factor plots, stability tables, and flags QC failures per run. The signed PDF includes every calibration curve, QC performance chart, IS-response trend, and full audit trail. [email protected] for a run on your targeted panel data.