BioinfoXpert

Bioinformatics Toolkit

ACP-Finder: Anticancer Peptide Prediction

Paste peptide sequences or upload FASTA files to predict anticancer potential directly in your browser. ACP-Finder combines amino acid composition, atomic signatures, and Shannon entropy with a trained random forest to deliver fast, privacy-preserving insights.

How ACP-Finder works

All computation happens in your browser. Sequences never leave your device.

Top sequence Probability 0.91 Classification ACPs

Predict anticancer peptides instantly

Upload FASTA files or paste peptide sequences to score anticancer potential using the trained ACP-Finder ensemble. The model evaluates amino acid composition, atomic content, and Shannon entropy to output a probability and class label for each sequence.

Input sequences

0 sequences ready

Results

Probabilities are averaged across 100 decision trees.

No predictions yet. Paste sequences and click Run prediction.

Model pipeline

How ACP-Finder works

Every prediction blends sequence statistics with physicochemical fingerprinting. ACP-Finder maps each peptide into a 26-feature space, then lets a 100-tree ensemble vote on anticancer potential.

  1. 1

    Sequence ingestion

    FASTA uploads and pasted peptides are normalised, deduplicated, and queued for feature extraction.

  2. 2

    Feature encoding

    SEP captures complexity, ATC measures elemental ratios, and AAC tracks residue enrichment. All client-side.

  3. 3

    Ensemble voting

    A random forest of 100 calibrated trees smooths predictions, yielding both class labels and ACP probabilities.

  4. 4

    Structure preview (optional)

    ACP-positive sequences can be sent to ESMFold for instant 3D inspection. No data leaves the page until you request it.

Interpreting the output

  • Probabilities > 0.5 Tip towards anticancer behaviour; we show the max-probability class per sequence.
  • Sequence length matters Extremely short peptides (<8 AA) may yield conservative scores. Consider validating borderline hits.
  • On-device inference No data leaves your browser; you can export results locally for downstream analysis.
  • Model updates Swap in a new acp-model.json to refresh predictions without touching the UI.
  • ESMFold only on demand Structure previews trigger remote calls only for ACP-labelled sequences when you ask for them.

Feature spotlight

Entropy (SEP) 0.0
ATC C% 0.0
AAC Lys% 0.0

Live values update after each prediction so you can see which signals dominate your sequences.

In-browser Anticancer Screening

About ACP-Finder

ACP-Finder (ACPF) evaluates the anticancer potential of peptide sequences directly in your browser. No Python runtimes, servers, or installs required. Upload FASTA files or paste sequences to generate calibrated probabilities and class labels in seconds.

Key capabilities

  • Batch scoring: process dozens of sequences in a single run via paste or FASTA upload.
  • Calibrated probabilities: a 100-tree random forest smooths predictions across diverse peptides.
  • Offline-first: feature extraction and inference stay on-device for private, reproducible analysis.
  • Export-ready: download neatly formatted CSV tables for downstream reporting.

Authorship

ACP-Finder is a collaboration between:

  • Naeem Mahmood Ashraf
  • Arslan Hamid
  • Kashaf Azam

Their work combines balanced training datasets with Pfeature-derived descriptors to achieve 89% accuracy on anticancer classification benchmarks.

  • Accuracy 89%
  • Trees 100
  • Features 26