usage.md

In gongyh/RamanD2O: RamanD2O: an R/Shiny app designed for the analysis of cellular Ramanome data

Usage example

RamanD2O supplies a small dataset for testing.

1. Load data

Just click "Quick start, click here to load a small demo dataset.", a small dataset with 25 spectra will be loaded. This dataset can be grouped using metadata columns "Timepoint" or "Label". There are three groups: I245_0mM, I245_0.5mM, and I245_1mM.

2. Trim wavelength

Raman shift of 450~3100 cm-1 are kept for subsequent analysis.

3. Spectra smoothing

Raman spectra are smoothed using SG method to increase SNR.

4. Baseline removal

Baselines are removed using als method.

5. Normalization

Raman spectra are normalized according the area of "Fingerprint" region.

6. SNR calculation

SNR of each Raman spectra is calculated using the "new" method. Low-SNR spectra with SNR<2.5 are removed.

7. CDR calculation

CDR of each spectra are calculated. As expected, all the spectra have D2O Raman peaks.

8. Visualization

The average spectra and CDR can be visualized and compared.

9. Multivariable statistics

This Raman spectra dataset can be further mined using multivariable statistics, e.g., PCA, LDA and MCR.

10. Machine learning

To test whether Fingerprint region can be used to separate these three groups, a random forest model is trained and evaluated.

gongyh/RamanD2O documentation built on Dec. 13, 2024, 8:39 a.m.