Host-cell proteins (HCPs) must be detected and quantified against the much higher abundance of the therapeutic antibody.
We reanalysed public data-independent acquisition (DIA) data for NISTmAb, a reference monoclonal antibody, prepared by native digestion or SP3 (single-pot, solid-phase-enhanced sample preparation). [1] We compared the DIA-NN results with identifications deposited by the original study.
HCPs detected in NISTmAb
Six injections per methodNative digestion
SP3 digestion
01 / DATA COMPLETENESS
HCP detection across preparations
Two NISTmAb digests were prepared independently by each method, and each digest was injected three times.
Detection frequency per HCP
Six injections per methodNative digestion 604 HCPs in total
in all six injections
SP3 digestion 506 HCPs in total
in all six injections
Within and between digests
For 241 of the 243 SP3 HCPs detected in exactly three injections, all three detections came from digest 2.
Detection in both digests was more common after native digestion: 64.7% of identified HCPs, compared with 36.6% after SP3.
| Detected in | DIA-NN groups | Deposited accessions |
|---|---|---|
| Digest 1 | 187 | 42 |
| Digest 2 | 504 | 88 |
| Both digests | 185 | 42 |
| All six injections | 9619.0% | 2831.8% |
02 / CHROMATOGRAMS
Peptide evidence for an HCP identification
We examined a peptide from protein disulfide-isomerase A6 (Pdia6; Q922R8) across the six native NISTmAb injections.
HQSLGGQYGVQGFPTIK3+
Relative intensity · 0–1 per panelTime from reported retention time (s) · dashed trace: MS1
Coeluting fragment ions and consistent relative intensities across all six injections support the peptide identification.
03 / QUANTUMS
Repeatability of HCP quantification
We assessed QuantUMS quantification using protein coefficients of variation (CVs) across three injections of each digest. [3]
Native digestion
15.9%median protein CV451 / 604 HCPs quantified in all three injections of at least one digest
SP3 digestion
20.7%median protein CV358 / 506 HCPs quantified in all three injections of at least one digest
Injection precision by digest
SP3 digest 1 yielded fewer HCP identifications and a higher median CV than digest 2.
285 of 451 native-digest HCPs and 174 of 358 SP3 HCPs had CVs below 20%.
| Preparation | HCPs | Median CV |
|---|---|---|
| Native · digest 1 | 448 | 14.7% |
| Native · digest 2 | 272 | 12.9% |
| SP3 · digest 1 | 90 | 30.0% |
| SP3 · digest 2 | 356 | 16.0% |
04 / HERCEPTIN
Antibody sequence coverage in Herceptin
Herceptin (trastuzumab, lot H4759H01-A025) was analysed in three SP3 digests, each injected twice. [1] The same heavy- and light-chain residues were covered in all six injections.
Heavy chain
65 distinct peptide sequences
387 of 450 residues
Light chain
25 distinct peptide sequences
177 of 214 residues
detected
The original study likewise reported no HCPs for this lot (Table S4DD). [1]
05 / METHODS
Analysis details
Samples and acquisition
We analysed 18 files from PXD067004. [1] NISTmAb native digestion and SP3 each comprised two independent digests with three injections per digest. Herceptin comprised three SP3 digests with two injections each.
Orbitrap Astral DIA data were acquired with a 19.9 min gradient in a 24 min LC method and nominal loading of 500 ng per injection. Native digestion used 1 mg NISTmAb; SP3 used 50 μg.
DIA-NN analysis
Each six-injection experiment was analysed separately with DIA-NN 2.7.0 Enterprise, using Generic scoring, MBR, Knowledge base and QuantUMS (precision). [2] [3] We used MS1 and MS2 mass tolerances of 4 and 10 ppm, respectively, and a calibration library derived from public K562 data. Proteotypicity was set to Isoform IDs.
Library prediction used Trypsin/P, up to two missed cleavages, peptide lengths of 7–30 residues, charges 1–4, precursor m/z 380–980 and fragment m/z 150–2000. Carbamidomethyl (C) was fixed; N-term M excision, Ox (M) and Acetyl (N-term) were enabled, with a maximum of one variable modification allowed.
Search databases contained reviewed mouse sequences (UniProt 2023_02) for NISTmAb or Chinese hamster sequences (UniProt 2026_03) for Herceptin, plus the corresponding antibody-chain sequences, the E. coli ClpB sequence and DIA-NN’s built-in contaminants.
HCP identification
To obtain HCP counts and calculate CVs, we applied 1% run-specific and global protein-group q-value filters (PG.Q.Value and Global.PG.Q.Value). We counted groups containing only mouse proteins for NISTmAb or Chinese hamster proteins for Herceptin.
For both DIA-NN and deposited HCP counts, we excluded antibody chains, ClpB, database contaminants, keratins, immunoglobulin-related proteins, desmosomal contaminants and proteins listed in the original study’s Tables S6/S7.
Protein quantification
Protein CVs were calculated within each digest using at least three positive quantities and averaged across digests.
Deposited results
We counted mouse protein accessions linked to peptide identifications in the deposited native and SP3 mzIdentML files. SP3 injection assignments came from the corresponding mzML file.
The analyses used different databases and processing settings. We did not independently validate the search-engine outputs.
Herceptin coverage
Herceptin sequence coverage was calculated using 1% run-specific and global precursor q-value filters (Q.Value and Global.Q.Value).
References
- Smith J et al. Quantitative Host Cell Protein Analysis of Antibody-Based Protein Therapeutics Using the Orbitrap Astral Mass Spectrometer. J Am Soc Mass Spectrom. 2026;37:1077–1094. doi:10.1021/jasms.5c00272. Data: PXD067004.
- Demichev V et al. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. Nat Methods. 2020. doi:10.1038/s41592-019-0638-x.
- Grossmann JL et al. Accurate quantification in proteomics with QuantUMS. Nat Biotechnol. 2026. doi:10.1038/s41587-026-03131-2.