Research Batch — How to Use

Use Research Batch when your scientific question depends on a series of FTIR spectra rather than one file: treatments, concentrations, formulations, time points or biological groups with replicates. The workflow keeps the study structure visible while you move from group patterns to the spectral regions behind them.

Choose this workflow forComparisons between experimental groups

Use it when replicates and independent samples matter to the scientific conclusion, not when you only need to identify one unknown spectrum.

The question it helps answerDo the groups differ, how consistent are they, and which bands explain the difference?

Similarity, PCA, clustering and peak differences are different views of the same study and should be interpreted together.

What you can carry forwardNumerical evidence for statistics, figures and manuscript review

Export values rather than relying on screenshots, while keeping sample assignment and spectral evidence available for co-authors or later analysis.

What this workbench does

Organizes many spectra into samples, replicates, and optional classes.
Compares spectra and isolate means with similarity metrics.
Runs PCA to show the main patterns of variation across the batch.
Builds UPGMA clustering to show relative sample relationships.
Highlights peaks that differ between defined groups.
Can build a SIMCA-style class model when there are at least 5 independent isolates per class.

When to use Research Batch

Formulation series, treatment groups, time points, or biological groups.
Replicate consistency checks within an experiment.
Exploratory pattern, grouping, and outlier review.
Finding bands that contribute to differences between predefined groups.

Important: Research Batch provides exploratory numerical evidence. Group separation in PCA or clustering is not, by itself, proof of chemical identity or causation.

The workflow in four steps

1Upload spectraUpload the FTIR files from your study. Multiple files can be added in one batch.

2Describe the samplesAssign each file to a sample, replicate or isolate, and optional class label. For larger batches, use the metadata CSV option.

3Run the analysisThe workbench calculates similarity, PCA, clustering, and peak differences using the selected settings.

4Review and exportInspect plots and tables, then export numerical results for independent statistics or reporting.

Sample and replicate assignment

Use sample IDs that make sense outside the software. Replicates should represent repeated measurements of the same sample; independent isolates should remain distinguishable when they represent separate experimental units.

Sample ID — A stable name for the sample or experimental unit.
Replicate / isolate — Use the assignment fields to keep repeated measurements and independent samples correctly grouped.
Class label — Optional predefined group such as Control, Treatment, or formulation group.

Analysis range and normalization

Fingerprint range: The default analysis range is 1800–900 cm⁻¹. Change it when your protocol requires a different region; peaks outside the selected range are not used in those calculations.

Normalization: Vector normalization is the default. SNV is also available when it is appropriate for your dataset. Keep preprocessing consistent across the spectra you intend to compare.

What the result tabs mean

Similarity Matrix — Pairwise comparison of spectra or isolate means using the available similarity or distance metrics.
PCA & SIMCA — PCA scores and explained variance; class-model diagnostics are shown when the data meet the class-model requirements.
Clustering — UPGMA dendrogram showing relative relationships in the analyzed feature space.
Peak Differences — Bands with the strongest group differences and the available statistical evidence.
Samples — The files and metadata used in the analysis.

SIMCA-style class model

The class model requires at least 5 independent isolates per class. It is a calibration model built from the data in the batch, not an independently validated classifier. Use external validation before treating a model as a predictive classification method.

Export results

Use CSV, Excel, or JSON exports to retain the numerical results and continue analysis outside FTIR.fun. The exported values are more suitable than screenshots for independent statistics and record keeping.

FAQ

How many spectra can I analyze at once?

Large batches take longer to process and can make plots or tables harder to read. Practical capacity depends on file size, the selected analysis, and server resources, so split very large studies into meaningful batches when needed.

What if I do not have replicates?

Treat each file as an independent sample when that matches your experimental design. Do not invent replicate relationships simply to satisfy an analysis requirement.

Can I compare spectra from different instruments?

Yes, but instrument, accessory, resolution, sampling, and preprocessing differences can create apparent group separation. Harmonize acquisition and preprocessing where possible and interpret cross-instrument differences cautiously.

Does a significant peak difference identify the chemical cause?

No. It identifies a spectral difference to investigate. Use chemical context, literature evidence, standards, or other analytical methods to support the assignment.