UniDL4BioPep: A universal deep learning architecture for bioactive peptide prediction

The webserver is the implementation of the paper "Du, Z., Ding, X., Xu, Y., & Li, Y. (2023). UniDL4BioPep: a universal deep learning architecture for binary classification in peptide bioactivity. Briefings in Bioinformatics, bbad135."

Notice: For very large dataset processing: please download our model locally or contact us at zhenjiao@ksu.edu or yonghui@ksu.edu for more assistant.

Result Explanation

sequence Antihypertensive Antihypertensive_probability

QKTAP active 0.981061339

NNWNG non-active 0.045210306

"Antihypertensive probability" is an indicator of how confident the model is with its prediction.

For example:

- A probability of 0.981061339 indicates the model is 98.1% confident in the active prediction.

- A probability of 0.045210306 indicates the model is only 4.5% confident in the active prediction, resulting in a non-active output.

Quick output version: 1. Choose a model → 2. Input a peptide sequence

Choose a model: Antihypertensive DPPIV Bitter Umami Antimicrobial Antimarial-alternative Antimalarial-main Quorum sensing Anticancer-alternative Anticancer-main Anti_MRSA Tumor T-cell antigens Blood-Brain Barrier Penetrating Peptides Anti-parasitic Neuropeptide Antibacterial Antifungal Antiviral Toxicity Antioxidant Allergen Cell Penetrating Peptide Run

Large-scale output version: 1. Prepare your files (xls, xlsx, fasta, or txt) and click “Choose File” for uploading → 2. Choose one or multiple models → 3. Download the results.

Choose your favorite model(s):
Antihypertensive DPPIV Bitter Umami Antimicrobial Antimarial-alternative Antimalarial-main Quorum sensing Anticancer-alternative Anticancer-main Anti_MRSA Tumor T-cell antigens Blood-Brain Barrier Penetrating Peptides Anti-parasitic Neuropeptide Antibacterial Antifungal Antiviral Toxicity Antioxidant Allergen Cell Penetrating Peptide
Choose a file to upload Run

Usage of the webserver:

Example for “Quick output version” :

1. Select “Antihypertensive” model for antihypertensive activity prediction.  →   →  →  2. Insert a peptide or protein sequence, “VPP” →   →  →  3. Click “Run”→   →  → 4. The result will be returned in seconds below the “Run” button

Notice: it also support multiple sequence at the same time. Just input as “VPP,IPP,CCL,AGR” (sequences are separated by comma, no space)

Example for “Large-scale output version:” :

1. Prepare your xls, xlsx, txt or fasta files → → → 2. Upload the file through “Choose File” botton → → → 3. Select one or several models → → → 4. Click “Run” → → → 5. It will automatically download your results.

Notice: File preparation should follow the examples under this repository https://github.com/dzjxzyd/UniDL4BioPep_webserver/tree/main/Example%20uploading%20files

Detailed explaination of the activity abbreviation

Antihypertensive: Angiotensin-converting enzyme inhibitory activity (main target in for hypertension); DPPIV: dipeptidyl peptidase IV (DPPIV) inhibitory activity (main target for diabetes); AMP: antimicrobial activity; AMAP: antimalarial activity (main and alternative is corresponding to two datasets; QS: quorum-sensing activity; ACP: anticancer activity: Anti_MRSA: anti-methicillin-resistant S. aureus strains activity); TTCA: tumor T cell antigens; BBP: blood-brain barrier penetrating; APP: anti-parasitic activity;

The whole model architecture

Dataset for the those models

Bioactivity	Training dataset	Test dataset
ACE inhibitory activity	913 positives and 913 negatives	386 positives and 386 negatives
DPP IV inhibitory activity	532 positives and 532 negatives	133 positives and 133 negatives
Bitter	256 positives and 256 negatives	64 positives and 64 negatives
Umami	112 positives and 241 negatives	28 positives and 61 negatives
Antimicrobial activity	3876 positives and 9552 negatives	2584 positives and 6369 negatives
Antimalarial activity	Main dataset (111 positives and 1708 negatives); alternative dataset (111 positives and 542 negatives)	Main dataset (28 positives and 427 negatives); alternative dataset (28 positives and 135 negatives)
Quorum sensing activity	200 positives and 200 negatives	20 positives and 20 negatives
Anticancer activity	Main dataset (689 positives and 689 negatives); alternative dataset (776 positives and 776 negatives)	Main dataset (172 positives and 172 negatives); alternative dataset (194 positives and 194 negatives)
Anti-MRSA strains activity	118 positives and 678 negatives	30 positives and 169 negatives
Tumor T cell antigens	470 positives and 318 negatives	122 positives and 75 negatives
Blood-Brain Barrier	100 positives and 100 negatives	19 positives and 19 negatives
Anti-parasitic activity	255 positives and 255 negatives	46 positives and 46 negatives
Neuropeptide	1940 positives and 1940 negatives	485 positives and 485 negatives
Antibacterial activity	6583 positives and 6583 negatives	1695 positives and 1695 negatives
Antifungal activity	778 positives and 778 negatives	215 positives and 215 negatives
Antiviral activity	2321 positives and 2321 negatives	623 positives and 623 negatives
Toxicity	1642 positives and 1642 negatives	290 positives and 290 negatives
Antioxidant activity	582 positives and 541 negatives	146 positives and 135 negatives