HPLC fingerprint chromatography, combined with rigorous data processing and chemometric analysis, has become the cornerstone of quality control for traditional Chinese medicine (TCM). This technical article explains the complete liquid chromatography fingerprint data processing workflow—from signal preprocessing and peak alignment to similarity evaluation and multivariate statistical modeling—supporting batch consistency, authentication, and regulatory compliance in line with pharmacopoeial, ISO, ICH, and GB standards.
Technical Overview
A chromatographic fingerprint is a characteristic, quantifiable chromatographic profile that represents the holistic chemical composition of a traditional Chinese medicine sample. Rather than quantifying one or two marker compounds, fingerprinting captures dozens to hundreds of simultaneous signals, making it ideally suited to the multi-component, multi-target nature of TCM products. High-performance liquid chromatography (HPLC) and ultra-high-performance liquid chromatography (UHPLC) are the dominant platforms for fingerprint generation, offering high separation efficiency, excellent reproducibility, and compatibility with UV, diode-array (DAD), evaporative light-scattering (ELSD), and mass spectrometric detectors.
The analytical value of a fingerprint, however, is determined by the quality of its data processing. Raw chromatograms inevitably contain baseline drift, detector noise, retention time shifts, and co-eluting peaks that distort inter-sample comparison. A systematic data processing pipeline—covering signal preprocessing, peak detection and matching, alignment, normalization, similarity evaluation, and chemometric pattern recognition—converts these raw signals into defensible, decision-ready metrics. Accredited third-party laboratories apply this pipeline within internationally recognized frameworks, including national pharmacopoeial technical requirements for TCM fingerprints, ISO standards developed for traditional Chinese medicine by ISO/TC 249, ICH Q2 analytical validation principles, and national standards of the GB series.
Testing Samples
Fingerprint data processing and analysis services cover the full range of sample types across the TCM value chain:
Crude and Processed Herbal Materials
Raw botanical and zoological drug substances, including sliced materials, processed (paozhi-treated) herbs, and samples collected from multiple geographical origins and harvest batches. Multi-origin datasets are essential for constructing reference fingerprint libraries and distinguishing authentic material from adulterants or substitutes.
Extracts and Intermediate Products
Concentrated extracts, purified fractions, resin-treated eluates, and granulation feedstocks. Fingerprint comparison at the intermediate stage enables early detection of process deviation before additional value is added downstream.
Finished TCM Products
Tablets, capsules, granules, oral liquids, tinctures, and injections. For injectable TCM products, fingerprint consistency requirements are particularly stringent, with regulatory guidance typically demanding high similarity values against an established reference fingerprint.
Reference Materials and Validation Samples
Chemical reference substances used for peak assignment and retention marking, together with replicate quality-control injections used to qualify system suitability, precision, and solution stability during the data acquisition campaign.
Testing Items
Typical data processing deliverables and evaluated parameters include:
- Retention time and relative retention time (RRT) tables for all detected peaks, referenced to a designated marker or reference peak
- Peak area and relative peak area (RPA) tables normalized against the reference peak or total signal
- Identification and matching of common (shared) peaks across the sample series, with matching rate statistics
- Cumulative proportion of total peak area represented by common peaks, typically required to exceed 80–90 percent
- Similarity values calculated as Pearson correlation coefficients and cosine-of-angle metrics against the consensus reference fingerprint
- Method validation metrics, including injection precision, reproducibility, sample solution stability over 24 hours, and system suitability
- Chemometric outputs such as PCA score and loading plots, HCA dendrograms, and validated PLS-DA/OPLS-DA discrimination models
- Peak assignment reports linking characteristic peaks to reference substances or tentative identities supported by MS data
Testing Methods
Chromatographic Data Acquisition and System Suitability
Fingerprints are acquired using validated HPLC or UHPLC methods with defined column chemistry, mobile phase composition, gradient profiles, flow rate, column temperature, and detection wavelength. System suitability is verified before batch analysis, with replicate injections typically required to achieve relative standard deviations of no more than 1 percent for retention time and no more than 3 percent for reference peak areas. Raw chromatographic data are exported into vendor-neutral formats such as AIA/ANDI (netCDF) to ensure transparent, software-independent processing.
Signal Preprocessing
Preprocessing removes systematic and random signal artifacts that would otherwise propagate into the statistical analysis. Baseline correction is performed using polynomial fitting or asymmetric least-squares regression, denoising and smoothing are applied through Savitzky–Golay filtering or wavelet transform methods, and solvent front or void-time regions are excluded from the analytical window. Preprocessing parameters are documented and fixed across all samples within a study to guarantee comparability.
Peak Detection, Integration, and Deconvolution
Peaks are detected using derivative-based or matched-filter algorithms with sensitivity and threshold settings qualified against signal-to-noise criteria. Automated integration is reviewed and corrected manually where necessary, and severely overlapping peaks are resolved using mathematical deconvolution techniques such as multivariate curve resolution or peak-purity assessment to ensure that integrated areas reflect true component responses.
Retention Time Alignment and Peak Matching
Small retention time shifts between injections are inevitable even under controlled conditions. Alignment is performed using marker-peak-based RRT correction, or advanced algorithms such as correlation optimized warping (COW) and dynamic time warping (DTW) when drift is more complex. Peaks are then matched across all chromatograms within defined time windows, producing a unified peak table in which each column represents one common peak across the entire sample set.
Normalization and Standardization
To compensate for variations in sample mass, extraction recovery, and injection volume, peak tables are normalized. Common approaches include normalization to total peak area, normalization to an internal standard or designated reference peak, and probabilistic quotient normalization. Prior to multivariate analysis, variables are additionally mean-centered and scaled, typically using unit variance (autoscaling) or Pareto scaling, to prevent large peaks from dominating the model.
Reference Fingerprint Generation and Similarity Evaluation
A consensus reference fingerprint is generated, usually as the median or mean chromatogram computed from a qualified set of representative batches. Similarity between each individual sample and the reference is expressed quantitatively using the Pearson correlation coefficient or the cosine of the angle between vectors. Acceptance thresholds are set by the applicable standard or product specification; finished products commonly require similarity values of 0.90 or higher, while broader limits may be justified for crude herbs of diverse origin.
Chemometric Pattern Recognition
Principal component analysis (PCA) is applied as an unsupervised method to visualize inherent clustering, detect outliers, and identify the variables driving separation. Hierarchical cluster analysis (HCA), typically using Ward linkage with Euclidean distance, classifies samples into groups. Where discrimination between predefined classes is required, supervised methods such as PLS-DA and OPLS-DA are used, supported by cross-validation, permutation testing, and examination of R2 and Q2 statistics to confirm that models are statistically sound and not overfitted.
Data Integrity and Reporting
All processing is conducted under data integrity principles consistent with ALCOA+ expectations, with audit trails, version-controlled processing parameters, and complete traceability from raw instrument files to final conclusions. Final reports include processed chromatogram overlays, common peak tables, similarity calculations, chemometric plots, and full method validation summaries in accordance with ICH Q2 principles.
Applications
Processed and analyzed fingerprint data support a wide spectrum of quality and regulatory objectives:
- Batch-to-batch consistency evaluation and routine release testing of herbal materials, extracts, and finished products
- Authentication of herbal species and discrimination of geographical origins, harvest years, and cultivation practices
- Detection of adulteration, substitution, and process shortcuts through deviation from established fingerprint patterns
- Stability studies supporting shelf-life determination, including accelerated and long-term storage condition comparisons
- Manufacturing process development, scale-up comparability, and continuous process verification
- Regulatory dossier support for product registration, where fingerprint consistency data are frequently requested by health authorities
- Research applications, including spectrum-effect relationship studies that correlate fingerprint variation with pharmacological activity
FAQ
What similarity value is considered acceptable for a TCM fingerprint?
Acceptance limits depend on the sample type and the governing standard. For finished products, and especially TCM injections, similarity values of at least 0.90 against the reference fingerprint are commonly required, while crude herbal materials of diverse origin may be assigned justified limits in the range of 0.80 to 0.90. The threshold must be defined in the product specification or validated study protocol.
How many common peaks are required in a valid fingerprint?
There is no universal number; the requirement is that the fingerprint be fully representative of the sample chemistry. Common peaks should generally account for more than 80 to 90 percent of the total detected peak area, and key characteristic peaks must be assigned with reference substances or supporting MS identification.
How is retention time drift between injections corrected?
Minor drift is corrected mathematically through relative retention time referencing to one or more stable marker peaks. For more complex shifts, warping algorithms such as correlation optimized warping or dynamic time warping are applied before peak matching, ensuring that the same compound is aligned in the same variable position across all samples.
Which detectors are suitable for fingerprint data processing?
DAD and UV detectors are most common and provide excellent reproducibility for chromophore-containing samples. ELSD is useful for components with weak UV absorption, such as sugars and saponins, while MS detection adds identification capability. Data from different detectors can be processed through the same alignment, normalization, and chemometric pipeline.
What deliverables are provided after fingerprint data processing?
Standard deliverables include raw and processed chromatogram overlays, unified common peak tables with RRT and RPA values, similarity evaluation results, PCA and HCA plots, validated supervised models where applicable, method validation summaries, and a conclusion on compliance with the defined fingerprint specification.
Can fingerprint data be correlated with biological activity?
Yes. By combining normalized fingerprint matrices with pharmacological datasets using correlation analysis, partial least squares regression, or network pharmacology approaches, characteristic peaks associated with efficacy can be identified, supporting the modernization and quality-by-design development of TCM products.
Why Choose ZKGX?
- State-of-the-art analytical equipment
- Highly qualified scientific team
- Fast turnaround time
- Competitive pricing