Build a PLS calibration from Raman spectra and known mRNA concentrations, then predict unknowns with 95% prediction intervals and outlier diagnostics. Runs entirely in your browser.
Try it out
Load example mRNA Raman data to see the full workflow
Drop SPC, JCAMP-DX, or two-column CSV/TSV files and enter the known concentration for each.
Pipeline: Savitzky–Golay smoothing → baseline correction (AsLS) → SNV → optional SG 1st derivative → PLS regression with leave-one-out CV.
Drop spectrum files of unknown concentration. Each prediction includes a 95% interval and an outlier flag based on Hotelling's T² and Q-residuals.
When to use
Do not use for
PLS extrapolates poorly. If you expect to measure 0.1–2 mg/mL, your calibration set should bracket that range. The 95% PI inflates outside the calibration centroid; the outlier flag warns you when an unknown is too far out.
The tool picks the LV count using a one-standard-error rule on RMSECV — the smallest LV count whose CV error is within 1 SE of the global minimum. If RMSECV keeps falling with more LVs, you're fitting noise. Two to four LVs is typical for clean Raman data.
An out-of-distribution score (high T²) means the spectrum sits in a part of latent-variable space the model didn't see. A high Q means there's structure in the spectrum the model can't explain. Either way, the prediction is an extrapolation.
PLS1 regression by NIPALS with leave-one-out cross-validation for latent-variable selection (one-standard-error rule). Preprocessing pipeline: Savitzky–Golay smoothing → asymmetric-least-squares baseline correction (λ = 10⁵, p = 0.01) → standard normal variate normalization → optional SG 1st derivative. Prediction interval at 95% from CV residual standard deviation, t-distributed with n−1 degrees of freedom, inflated by sample-specific Hotelling T². Outlier limits use the F-distribution-based Hotelling T² 95% bound and a max+2σ Q-residual bound calibrated on the training set.
Last validated 2026-04-25. Calculations are designed for planning and documentation support; verify procurement decisions against manufacturer specifications or institutional SOPs.
ConductScience. (2026). mRNA Raman Calibrator (v1.0) [Web tool]. https://conductscience.com/tools/mrna-raman-calibrator
Wold S, Sjöström M, Eriksson L. PLS-regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems. 2001;58:109–130. doi:10.1016/S0169-7439(01)00155-1
Eilers PHC, Boelens HFM. Baseline correction with asymmetric least squares smoothing. Leiden University Medical Centre Report. 2005.
Savitzky A, Golay MJE. Smoothing and differentiation of data by simplified least squares procedures. Analytical Chemistry. 1964;36:1627–1639. doi:10.1021/ac60214a047
Raman spectra of mRNA in aqueous solution are dominated by a small number of vibrational bands.
Intensity at the phosphate band (~1080) tracks total nucleic acid concentration well over the 0.05–5 mg/mL range typical of IVT and LNP work. Combining several bands in a multivariate model (PLS) outperforms univariate calibration, especially when buffers contribute background bands.
Partial Least Squares regression solves the same problem as linear regression — predicting concentration from spectrum — but copes with the fact that adjacent wavenumbers are highly correlated.
