Halal Origin Pre-screening · Guide
Pre-screening · FTIR.fun
Guide Sign in
Open Workbench
Guide

Halal Origin Pre-screening — How to Use

A step-by-step guide for building your private reference library and screening incoming samples for porcine-origin tendency. Reading time: about 5 minutes.

What this workbench does

  • Keeps a private, per-team reference library of FTIR spectra from qualified batches you trust.
  • Compares a new incoming sample against that library and reports how consistent it is, with peak-level difference evidence.
  • Works for any category: protein-based (gelatin, capsule shell, meat, surimi, dairy), fat/oil-based (edible oil, shortening, butter, ghee), or mixed processed products.
  • Gives you a traceable report you can print and keep with the batch records.

How the result is produced

  • The full spectrum is analysed — the literature windows shown on the page are reading guides, not analysis limits.
  • The conclusion is a similarity judgement: how much does the incoming sample match your qualified reference library.
  • Result wording follows your library tags: a non-porcine library gives non-porcine wording, a porcine control library gives porcine wording.
  • Every label is backed by the peak-difference table you can review in the report.

The workflow in four steps

1Create a reference libraryUpload qualified historical FTIR batches for one material.
2Tag each batchMatrix type and source species decide how results are worded.
3Screen an incoming sampleUpload its FTIR file against the selected library.
4Read the reportTendency label, peak differences, charts, and AI interpretation.

Step 1 — Create a reference library

A reference library is a set of FTIR spectra from batches you already consider acceptable for this material. The incoming sample is compared against this library, not against any public database.

What files can you upload?

Common instrument export formats are accepted, including CSV / TXT (wavenumber, absorbance), SPA, DPT, DXD, JWS, and others. The spectrum is read and interpolated automatically; you do not need to convert it by hand.

How many batches do you need?

Start with at least 3–5 batches for one material. More batches make the comparison more reliable. The page shows a warning whenever the selected library has fewer than 5 batches.

One library per material

Create a separate library for each material you want to screen (e.g. one for beef sausage, one for cooking oil). Libraries are private to your team account and appear in the left sidebar.

Matrix type

Choose Protein-based, Fat/oil-based, or Mixed. This only changes which literature discriminant windows are shown on the page as reading guides — the analysis itself always uses the full spectrum.

Source species — required, and why

Tag each batch with its declared source (bovine, fish, plant, porcine, ...). The result wording depends on this: a library of certified non-porcine batches produces “non-porcine tendency” wording, while a library tagged as a porcine positive control produces “porcine tendency” wording on a close match. Without species tags, the report only states library consistency.

Other fields

Supplier, batch number, JAKIM/cert status, scans, resolution, and ATR crystal are recorded on every reference batch and appear in reports, so the traceability chain stays intact.

Step 2 — Screen an incoming sample

  1. Select the reference library for this material in the “Reference library” dropdown.
  2. Enter the batch number and the supplier-declared species of the incoming sample (this is recorded in the report, not used by the engine).
  3. Upload the incoming FTIR file (same formats as above).
  4. Click “Run Halal Pre-screening”. The engine compares the sample against the library and produces the report.
Points: each pre-screening consumes 10 points on a free/trial account. VIP accounts are not charged.

Step 3 — Read the report

The tendency label

The seal at the top is the pre-screening conclusion. What it can say depends on what your reference library is:

Your library Sample vs library Label
Non-porcine certified batchesConsistentConsistent with non-porcine references
Partly consistentNeeds further confirmation
InconsistentNon-porcine origin not confirmed
Porcine positive controlConsistentPorcine tendency
Partly consistentNeeds further confirmation
InconsistentDoes not match the porcine control
No species tagConsistentConsistent with the reference library
Partly consistentNeeds further confirmation
InconsistentInconsistent with the reference library

The algorithm behind the label

The engine automatically picks the analysis mode by how many batches the library has:

Reference batchesModeWhat is calculated
19 or fewerSimilarityPearson correlation, cosine similarity, and the main difference peaks
20 – 49PCASimilarity plus PCA projection, Mahalanobis distance, Hotelling T²
50 or moreClass modelPCA plus SIMCA: Q residual, class membership probability, conformity score

The discriminant regions

The “Discriminant regions (literature)” panel lists windows reported in published origin-screening studies, matched to your matrix type. Peaks in the difference table that fall inside these windows are tagged with the region name — a useful prioritisation for your own review:

Amide A 3600–3000 cm⁻¹ N–H/O–H stretch; principal gelatin discriminant window (Hashim 2010).
Amide I 1700–1600 cm⁻¹ C=O stretch; porcine vs bovine/fish gelatin main peak (Sci Rep 2021).
Amide II / III 1550–1240 cm⁻¹ N–H bend + C–N stretch; species-discriminant envelope (Sci Rep 2021).
Porcine-gelatin signature 1475–1470 cm⁻¹ Dominant wavenumber reported for porcine gelatin by c-FACS (Sci Rep 2021); bovine 1444–1450, fish 1496–1500.
DA window 2000–1501 cm⁻¹ Standard ATR + discriminant-analysis window in animal-gelatin studies (Nazri 2024).
C–H stretch 3030–2785 cm⁻¹ Total fat C–H region; SIMCA discriminant region for lard in ghee (Int J Mol Sci 2020).
Ester C=O 1786–1680 cm⁻¹ Triglyceride ester carbonyl; SIMCA region separating lard from other fats (2020 review).
Triglyceride fingerprint 1490–919 cm⁻¹ SIMCA discriminant region (2020 review); lard C–O pair reported near 1117/1097.
Lard in VCO region 3020–3000 · 1120–1000 cm⁻¹ Combined region that quantified lard in virgin coconut oil by PLSR+DA, R²=0.999 (2020 review).
cis / trans markers 3006 · 968 cm⁻¹ cis =C–H near 3006 and trans/CH₂ rocking near 968 support fat-type separation.

Where these windows come from

Protein-based windows follow Hashim 2010 (Food Chem) and Sci Rep 2021 — the latter reports a dominant wavenumber for porcine gelatin near 1470–1475 cm⁻¹ (bovine 1444–1450, fish 1496–1500). Fat/oil-based windows follow the Int J Mol Sci 2020 halal-authentication review: SIMCA discriminant regions for lard in ghee (3030–2785 / 1786–1680 / 1490–919 cm⁻¹) and a combined region that quantified lard in virgin coconut oil by PLSR+DA with R² = 0.999 (3020–3000 / 1120–1000 cm⁻¹).

These are reading guides for your own review — the engine analyses the whole spectrum, not only these windows. There is no single window set valid for every material; different matrices have different literature discriminant bands.

The peak difference table

“Main Difference Peaks” lists the strongest differences between the sample and the reference mean, with the direction (sample higher/lower) and whether the peak is anomalous. This is the evidence behind the label — review it when the label is not “consistent”.

Charts

The spectral overlay draws sample vs reference mean. The PCA scatter (when the library has 20+ batches) shows where the sample sits relative to the historical batch cloud.

AI interpretation

When available, the AI panel summarises the differences, lists plausible causes, and suggests next steps. It is decision-support only: treat its chemical claims and wording as a starting point, and verify before acting.

FAQ

My sample came back “non-porcine origin not confirmed”. Does that mean it contains pork?
No. It means the sample differs from your non-porcine reference library. Differences can come from a different supplier, moisture, additives, instrument settings, or normal batch-to-batch spread. Treat it as a flag to investigate — not as a porcine detection. If porcine presence must be established, use DNA/PCR.
Can I screen oils and fats with this?
Yes. Choose “Fat / oil-based” as the matrix type. The literature panels then show the bands reported for lard adulteration screening (cis C–H near 3006, ester C=O near 1743, the triglyceride C–O pair near 1117/1097, trans near 968), and the AI interpretation gives fat-specific reading guidance.
Why does the page warn that results are exploratory with fewer than 5 batches?
A comparison against 1–4 spectra has limited statistical meaning. The engine still works, but treat the output as directional, not quantitative. Add batches as you qualify them.
Is my spectral data shared?
No. Reference libraries are private to your account and are not used as testing references for anyone else. You can remove batches or whole libraries at any time.
Do I need to create one library per supplier or per material?
Per material is the minimum. If your suppliers differ strongly, separate libraries per supplier per material give the cleanest comparison.
How is this different from a similarity search in a big spectral library?
Here the comparison set is your own qualified batches, the evidence is peak-level differences with literature windows, and the report keeps the supplier/batch/cert trail. A public-library search can answer “what is this spectrum similar to”; this workbench answers “how much does this batch match my qualified baseline”.
Can I turn a screened sample into a reference batch?
Yes — on the workbench, each past screening has an “Add” button that promotes it into the reference library once you accept the batch. Records keep their history.
What does “JAKIM status” mean for users outside Malaysia?
The field is free-form and labelled “JAKIM / halal cert status”. Laboratories outside Malaysia can record their own certification body, or leave it as “Unknown”.