🧬 1. Generation Module

Overview

The Generation module uses deep learning models to generate novel antimicrobial peptides (AMPs) with specific properties based on your target pathogen selection. Generated peptides are automatically evaluated for their antimicrobial activity, safety profile, and overall therapeutic potential.

How to Use

1. Navigate to the Generation Page
   Click on "Generation" in the navigation menu.
2. Select Target Pathogen
   Choose one of the following options:

• Broad-spectrum AMPs: Generates peptides active against both Gram-positive and Gram-negative bacteria
• Gram-negative-specific AMPs: Generates peptides targeting Gram-negative bacteria (e.g., Escherichia coli)
• Gram-positive-specific AMPs: Generates peptides targeting Gram-positive bacteria (e.g., Staphylococcus aureus)

1. Set Number of Sequences
   Specify how many peptide sequences you want to generate (typically 10-100 sequences recommended).
2. Submit Job
   Click the "Generate" button to start the generation process. You'll receive a unique Job ID.
3. Track Progress
   Use the Job ID to monitor your job in the Job Status page or wait on the current page for results.

Example Output

Here's what the generated results look like:

Download Results

Once the job is complete, you can download the results as a CSV file containing all generated sequences and their evaluation scores. The file name will include your Job ID for easy reference.

🔬 2. Screening Module

Overview

The Screening module evaluates your own peptide sequences using our trained models. Upload a list of peptide sequences, and the system will predict their antimicrobial properties, antibacterial activity score, and hemolytic score.

How to Use

1. Navigate to the Screening Page
   Click on "Screening" in the navigation menu.
2. Prepare Your Sequences
   Format your peptide sequences as follows:
   >peptide_1 GLFDIVKKVVGAFGSL >peptide_2 KWKLFKKIGIGAVLKVL >peptide_3 FFHHIFRGIVHVGKTIHRLVTG

• FASTA format with header lines starting with ">"
• One sequence per peptide
• Amino acid sequences using standard single-letter codes
• Maximum file size: 5MB

1. Upload Sequences
   You can either:

• Paste sequences directly into the text area, or
• Upload a FASTA file (.fasta, .fa, .txt)

1. Submit Job
   Click "Screen Sequences" to start the analysis. You'll receive a unique Job ID.
2. Review Results
   Results are sorted by score (highest to lowest) to help you identify the best candidates quickly.

Example Input

Example Output

Here's what the screening results look like:

Download Results

Download results as CSV for further analysis in Excel, R, Python, or other tools. Results include all original sequence names for easy tracking.

📊 3. Job Status Module

Overview

The Job Status module allows you to track the progress of your submitted jobs, view completed results, and download output files.

How to Use

1. Navigate to Job Status Page
   Click on "Job Status" in the navigation menu.
2. Enter Your Job ID
   Paste the Job ID you received when submitting your generation or screening job.
   Example Job ID: 9a972711-7516-4d61-a30d-c45731daa5b1
3. Check Status
   Click "Check Status" to view your job information. You'll see:

• Status: PENDING, PROCESSING, COMPLETED, or FAILED
• Progress: Percentage complete (if available)
• Type: Generation or Screening
• Created: Submission timestamp

1. Download Results
   Click the "Download Results" button to save the complete CSV file to your computer.

🎯 4. Scoring System Explained

Overview

Our scoring system evaluates peptides based on multiple criteria to predict their potential as therapeutic antimicrobial agents. Each criterion contributes to an overall score.

Scoring Criteria

1. AMP Score (Antimicrobial Peptide Identification)

What it measures: The probability that a peptide is an antimicrobial peptide.
Threshold: 0.998
Scoring: Score > 0.998 → +1 point
Model: PeptideBERT-LSTM trained on curated AMP databases

2. E.coli Activity Score (Gram-negative Activity)

What it measures: Probability of antimicrobial activity against E.coli (Gram-negative bacteria).
Threshold: 0.9980798959732056
Scoring: Score > threshold → +1 point (active against E.coli)
Note: Only evaluated for broad-spectrum and E.coli-specific generations

3. S.aureus Activity Score (Gram-positive Activity)

What it measures: Probability of antimicrobial activity against S.aureus (Gram-positive bacteria).
Threshold: 0.9989857077598572
Scoring: Score > threshold → +1 point (active against S.aureus)
Note: Only evaluated for broad-spectrum and S.aureus-specific generations

4. Hemolytic Score (Safety Assessment)

What it measures: Probability of hemolytic activity (damage to red blood cells).
Threshold: 0.5
Scoring: Score ≤ 0.5 → +1 point (non-hemolytic, safe)
Importance: Critical for drug safety - hemolytic peptides are toxic

Overall Evaluation

For Screening & Broad-spectrum Generation (4-point scale):

For Target-Specific Generation (3-point scale):

🤖 5. Model Details & Thresholds

Model Architecture

AIPAMPDS uses state-of-the-art deep learning models based on transformer architectures and recurrent neural networks, trained on large-scale antimicrobial peptide databases.

Threshold Selection

Model thresholds are optimized to balance sensitivity and specificity:

• High thresholds (e.g., 0.998) ensure high confidence predictions
• Validated on experimental data from peer-reviewed publications
• Minimize false positives to reduce wasted experimental effort

Training Data

❓ 6. Frequently Asked Questions

Q: How long are my results stored?

Results are stored on our servers for 30 days. After this period, they may be automatically deleted to free up space. We recommend downloading your results promptly after job completion.

Q: Can I use the generated sequences commercially?

Yes, you can use the generated sequences for research and commercial purposes. However, we recommend:

• Experimental validation before any application
• Patent searches to ensure novelty
• Consulting with legal experts for IP protection

Q: What if my job fails?

Job failures can occur due to:

• Invalid input format (check FASTA formatting)
• Server issues (temporary - try again later)
• Sequences too long (>100 amino acids may cause issues)

Q: How accurate are the predictions?

Our models achieve >95% accuracy on validation datasets. However:

• Predictions are computational estimates, not guarantees
• Experimental validation is essential
• Novel sequences may behave differently than training data
• Always test in lab conditions before applications

Q: Can I generate very long peptides?

Our generation models are optimized for peptides of 10-40 amino acids. While longer sequences can be screened, generation quality may decrease for very long peptides (>50 amino acids).

Q: What does "Broad-spectrum" mean?

Broad-spectrum antimicrobials are active against both Gram-positive and Gram-negative bacteria. This is valuable for treating infections with unknown or mixed bacterial populations.

Q: Why are some scores very close to the threshold?

Scores near the threshold (e.g., 0.9979 vs threshold 0.998) indicate borderline cases. These peptides may have marginal activity and should be interpreted with caution. We recommend focusing on peptides with scores significantly above thresholds.

Q: Can I request new features or models?

Yes! We welcome feedback and feature requests. Please use the Contact page to suggest new pathogens, additional properties to predict, or other enhancements.