app.py

import gradio as gr
import torch
import joblib
import numpy as np
from itertools import product
import torch.nn as nn
import matplotlib.pyplot as plt
import io
from PIL import Image

class VirusClassifier(nn.Module):
    def __init__(self, input_shape: int):
        super(VirusClassifier, self).__init__()
        self.network = nn.Sequential(
            nn.Linear(input_shape, 64),
            nn.GELU(),
            nn.BatchNorm1d(64),
            nn.Dropout(0.3),
            nn.Linear(64, 32),
            nn.GELU(),
            nn.BatchNorm1d(32),
            nn.Dropout(0.3),
            nn.Linear(32, 32),
            nn.GELU(),
            nn.Linear(32, 2)
        )

    def forward(self, x):
        return self.network(x)

def parse_fasta(text):
    """Parse FASTA formatted text into a list of (header, sequence)."""
    sequences = []
    current_header = None
    current_sequence = []
    
    for line in text.strip().split('\n'):
        line = line.strip()
        if not line:
            continue
        if line.startswith('>'):
            if current_header:
                sequences.append((current_header, ''.join(current_sequence)))
            current_header = line[1:]
            current_sequence = []
        else:
            current_sequence.append(line.upper())
    if current_header:
        sequences.append((current_header, ''.join(current_sequence)))
    return sequences

def sequence_to_kmer_vector(sequence: str, k: int = 4) -> np.ndarray:
    """Convert a sequence to a k-mer frequency vector."""
    kmers = [''.join(p) for p in product("ACGT", repeat=k)]
    kmer_dict = {km: i for i, km in enumerate(kmers)}
    vec = np.zeros(len(kmers), dtype=np.float32)
    
    for i in range(len(sequence) - k + 1):
        kmer = sequence[i:i+k]
        if kmer in kmer_dict:
            vec[kmer_dict[kmer]] += 1

    total_kmers = len(sequence) - k + 1
    if total_kmers > 0:
        vec = vec / total_kmers

    return vec

def calculate_shap_values(model, x_tensor):
    """
    Calculate SHAP values using a simple ablation approach.
    Returns shap values and model prediction.
    """
    model.eval()
    with torch.no_grad():
        # Get baseline prediction
        baseline_output = model(x_tensor)
        baseline_probs = torch.softmax(baseline_output, dim=1)
        baseline_prob = baseline_probs[0, 1].item()  # Probability of human class
        
        # Calculate impact of zeroing each feature
        shap_values = []
        x_zeroed = x_tensor.clone()
        for i in range(x_tensor.shape[1]):
            x_zeroed[0, i] = 0
            output = model(x_zeroed)
            probs = torch.softmax(output, dim=1)
            prob = probs[0, 1].item()
            impact = baseline_prob - prob  # How much removing the feature changed the prediction
            shap_values.append(impact)
            x_zeroed[0, i] = x_tensor[0, i]  # Restore the original value
            
    return np.array(shap_values), baseline_prob

def create_importance_bar_plot(shap_values, kmers, top_k=10):
    """Create a bar plot of the most important k-mers."""
    plt.rcParams.update({'font.size': 10})
    plt.figure(figsize=(10, 6))
    
    # Sort by absolute importance
    indices = np.argsort(np.abs(shap_values))[-top_k:]
    values = shap_values[indices]
    features = [kmers[i] for i in indices]
    
    colors = ['#ff9999' if v > 0 else '#99ccff' for v in values]
    
    plt.barh(range(len(values)), values, color=colors)
    plt.yticks(range(len(values)), features)
    plt.xlabel('SHAP value (impact on model output)')
    plt.title(f'Top {top_k} Most Influential k-mers')
    plt.gca().invert_yaxis()  # Most important at top
    
    return plt.gcf()

def visualize_sequence_impacts(sequence, kmers, shap_values, base_prob):
    """
    Create a SHAP-style visualization of sequence impacts.
    Shows each k-mer's contribution in context.
    """
    k = 4  # k-mer size
    kmer_dict = {km: i for i, km in enumerate(kmers)}
    
    # Find all k-mers and their impacts
    kmer_impacts = []
    for i in range(len(sequence) - k + 1):
        kmer = sequence[i:i+k]
        if kmer in kmer_dict:
            impact = shap_values[kmer_dict[kmer]]
            kmer_impacts.append((i, kmer, impact))
    
    # Sort by absolute impact
    kmer_impacts.sort(key=lambda x: abs(x[2]), reverse=True)
    
    # Create the plot
    fig = plt.figure(figsize=(20, max(10, len(kmer_impacts[:30])*0.3)))
    ax = plt.gca()
    
    # Add title and base value
    plt.text(0.01, 1.02, f"base value = {base_prob:.3f}", transform=ax.transAxes, fontsize=12)
    
    # Plot k-mers
    y_position = 1
    sequence_length = len(sequence)
    
    for pos, kmer, impact in kmer_impacts[:30]:  # Show top 30 most impactful k-mers
        # Show sequence with highlighted k-mer
        pre_sequence = sequence[:pos]
        post_sequence = sequence[pos+k:]
        
        # Choose color based on impact
        color = '#ffcccb' if impact > 0 else '#cce0ff'  # Light red or light blue
        arrow = '↑' if impact > 0 else '↓'
        
        # Calculate text positions
        plt.text(0.01, y_position, pre_sequence, fontsize=10)
        plt.text(0.01 + len(pre_sequence)/(sequence_length*1.5), y_position, 
                kmer, fontsize=10, bbox=dict(facecolor=color, alpha=0.3, pad=2))
        plt.text(0.01 + (len(pre_sequence) + len(kmer))/(sequence_length*1.5), 
                y_position, post_sequence, fontsize=10)
        
        # Add impact value
        plt.text(0.8, y_position, f"{arrow} {impact:+.3f}", fontsize=10)
        
        y_position -= 0.03
    
    plt.axis('off')
    plt.tight_layout()
    return fig

def predict(file_obj, top_kmers=10, fasta_text=""):
    """Main prediction function for Gradio interface."""
    # Handle input
    if fasta_text.strip():
        text = fasta_text.strip()
    elif file_obj is not None:
        try:
            with open(file_obj, 'r') as f:
                text = f.read()
        except Exception as e:
            return f"Error reading file: {str(e)}", None, None
    else:
        return "Please provide a FASTA sequence.", None, None

    # Parse FASTA
    sequences = parse_fasta(text)
    if not sequences:
        return "No valid FASTA sequences found.", None, None
    
    header, seq = sequences[0]

    # Load model and process sequence
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    try:
        model = VirusClassifier(256).to(device)
        model.load_state_dict(torch.load('model.pt', map_location=device, weights_only=True))
        scaler = joblib.load('scaler.pkl')
    except Exception as e:
        return f"Error loading model: {str(e)}", None, None

    # Generate features
    freq_vector = sequence_to_kmer_vector(seq)
    scaled_vector = scaler.transform(freq_vector.reshape(1, -1))
    x_tensor = torch.FloatTensor(scaled_vector).to(device)

    # Calculate SHAP values and get prediction
    shap_values, prob_human = calculate_shap_values(model, x_tensor)
    
    # Generate result text
    results = [
        f"Sequence: {header}",
        f"Prediction: {'Human' if prob_human > 0.5 else 'Non-human'} Origin",
        f"Confidence: {max(prob_human, 1-prob_human):.3f}",
        f"Human Probability: {prob_human:.3f}",
        "\nTop Contributing k-mers:"
    ]

    # Get k-mers for visualization
    kmers = [''.join(p) for p in product("ACGT", repeat=4)]
    
    # Create visualizations
    importance_plot = create_importance_bar_plot(shap_values, kmers, top_kmers)
    sequence_plot = visualize_sequence_impacts(seq, kmers, shap_values, prob_human)
    
    # Convert plots to images
    def fig_to_image(fig):
        buf = io.BytesIO()
        fig.savefig(buf, format='png', bbox_inches='tight', dpi=150)
        buf.seek(0)
        img = Image.open(buf)
        plt.close(fig)
        return img

    return "\n".join(results), fig_to_image(importance_plot), fig_to_image(sequence_plot)

# Create Gradio interface
css = """
.gradio-container {
    font-family: 'IBM Plex Sans', sans-serif;
}
"""

with gr.Blocks(css=css) as iface:
    gr.Markdown("""
    # Virus Host Classifier
    Predicts whether a viral sequence is of human or non-human origin using k-mer analysis.
    """)
    
    with gr.Row():
        with gr.Column(scale=1):
            file_input = gr.File(
                label="Upload FASTA file",
                file_types=[".fasta", ".fa", ".txt"],
                type="filepath"
            )
            text_input = gr.Textbox(
                label="Or paste FASTA sequence",
                placeholder=">sequence_name\nACGTACGT...",
                lines=5
            )
            top_k = gr.Slider(
                minimum=5,
                maximum=30,
                value=10,
                step=1,
                label="Number of top k-mers to display"
            )
            submit_btn = gr.Button("Analyze Sequence", variant="primary")
            
        with gr.Column(scale=2):
            results = gr.Textbox(label="Analysis Results", lines=10)
            kmer_plot = gr.Image(label="K-mer Importance Plot")
            shap_plot = gr.Image(label="Sequence Impact Visualization (SHAP-style)")
    
    submit_btn.click(
        predict,
        inputs=[file_input, top_k, text_input],
        outputs=[results, kmer_plot, shap_plot]
    )
    
    gr.Markdown("""
    ### Visualization Guide
    - **K-mer Importance Plot**: Shows the most influential k-mers and their SHAP values
    - **Sequence Impact Visualization**: Shows the sequence with highlighted k-mers:
      - Red highlights = pushing toward human origin
      - Blue highlights = pushing toward non-human origin
      - Arrows (↑/↓) show impact direction
      - Values show impact magnitude
    """)

if __name__ == "__main__":
    iface.launch()