app.py

import os
os.system("pip install streamlit pandas xlsxwriter openpyxl pymongo")

import streamlit as st
import pandas as pd
import xlsxwriter
from io import BytesIO
from collections import defaultdict
from pymongo import MongoClient
import hashlib

# MongoDB setup
client = MongoClient("mongodb+srv://dhruvmangroliya:[email protected]/BTP_DB?retryWrites=true&w=majority")
db = client['BTP_DB']
results_collection = db['protein_results']

# Utility
def is_homo_repeat(s):
    return all(c == s[0] for c in s)

def hash_sequence(sequence):
    return hashlib.md5(sequence.encode()).hexdigest()

@st.cache_data(show_spinner=False)
def fragment_protein_sequence(sequence, max_length=1000):
    return [sequence[i:i+max_length] for i in range(0, len(sequence), max_length)]

def find_homorepeats(protein):
    n = len(protein)
    freq = defaultdict(int)
    i = 0
    while i < n:
        curr = protein[i]
        repeat = ""
        while i < n and curr == protein[i]:
            repeat += protein[i]
            i += 1
        if len(repeat) > 1:
            freq[repeat] += 1
    return freq

def find_hetero_amino_acid_repeats(sequence):
    repeat_counts = defaultdict(int)
    for length in range(2, len(sequence) + 1):
        for i in range(len(sequence) - length + 1):
            substring = sequence[i:i+length]
            repeat_counts[substring] += 1
    return {k: v for k, v in repeat_counts.items() if v > 1}

def check_boundary_repeats(fragments, final_repeats, overlap=50):
    for i in range(len(fragments) - 1):
        left_overlap = fragments[i][-overlap:]
        right_overlap = fragments[i + 1][:overlap]
        overlap_region = left_overlap + right_overlap
        boundary_repeats = find_hetero_amino_acid_repeats(overlap_region)
        for substring, count in boundary_repeats.items():
            if any(aa in left_overlap for aa in substring) and any(aa in right_overlap for aa in substring):
                final_repeats[substring] += count
    return final_repeats

def find_new_boundary_repeats(fragments, final_repeats, overlap=50):
    new_repeats = defaultdict(int)
    for i in range(len(fragments) - 1):
        left_overlap = fragments[i][-overlap:]
        right_overlap = fragments[i + 1][:overlap]
        overlap_region = left_overlap + right_overlap
        boundary_repeats = find_hetero_amino_acid_repeats(overlap_region)
        for substring, count in boundary_repeats.items():
            if any(aa in left_overlap for aa in substring) and any(aa in right_overlap for aa in substring):
                if substring not in final_repeats:
                    new_repeats[substring] += count
    return new_repeats

def get_or_process_sequence(sequence, analysis_type, overlap=50):
    sequence_hash = hash_sequence(sequence)
    cached = results_collection.find_one({"_id": sequence_hash, "analysis_type": analysis_type})
    if cached:
        return cached["repeats"]

    fragments = fragment_protein_sequence(sequence)
    final_repeats = defaultdict(int)

    if analysis_type == "Hetero":
        for fragment in fragments:
            fragment_repeats = find_hetero_amino_acid_repeats(fragment)
            for k, v in fragment_repeats.items():
                final_repeats[k] += v
        final_repeats = check_boundary_repeats(fragments, final_repeats, overlap)
        new_repeats = find_new_boundary_repeats(fragments, final_repeats, overlap)
        for k, v in new_repeats.items():
            final_repeats[k] += v
        final_repeats = {k: v for k, v in final_repeats.items() if not is_homo_repeat(k)}

    elif analysis_type == "Homo":
        final_repeats = find_homorepeats(sequence)

    elif analysis_type == "Both":
        hetero_repeats = defaultdict(int)
        for fragment in fragments:
            fragment_repeats = find_hetero_amino_acid_repeats(fragment)
            for k, v in fragment_repeats.items():
                hetero_repeats[k] += v
        hetero_repeats = check_boundary_repeats(fragments, hetero_repeats, overlap)
        new_repeats = find_new_boundary_repeats(fragments, hetero_repeats, overlap)
        for k, v in new_repeats.items():
            hetero_repeats[k] += v
        hetero_repeats = {k: v for k, v in hetero_repeats.items() if not is_homo_repeat(k)}

        homo_repeats = find_homorepeats(sequence)
        final_repeats = homo_repeats.copy()
        for k, v in hetero_repeats.items():
            final_repeats[k] += v

    results_collection.insert_one({
        "_id": sequence_hash,
        "analysis_type": analysis_type,
        "repeats": dict(final_repeats)
    })
    return final_repeats

def process_excel(excel_data, analysis_type):
    repeats = set()
    sequence_data = []
    count = 0
    for sheet_name in excel_data.sheet_names:
        df = excel_data.parse(sheet_name)
        if len(df.columns) < 3:
            st.error(f"Error: The sheet '{sheet_name}' must have at least three columns: ID, Protein Name, Sequence")
            return None, None
        for _, row in df.iterrows():
            entry_id = str(row[0])
            protein_name = str(row[1])
            sequence = str(row[2]).replace('"', '').replace(' ', '').strip()
            if not sequence:  # Skip empty sequence
                continue
            count += 1
            freq = get_or_process_sequence(sequence, analysis_type)
            sequence_data.append((entry_id, protein_name, freq))
            repeats.update(freq.keys())
    st.toast(f"{count} sequences processed.")
    return repeats, sequence_data

def create_excel(sequences_data, repeats, filenames):
    output = BytesIO()
    workbook = xlsxwriter.Workbook(output, {'in_memory': True})
    for file_index, file_data in enumerate(sequences_data):
        filename = filenames[file_index]
        worksheet = workbook.add_worksheet(filename[:31])
        worksheet.write(0, 0, "Entry")
        worksheet.write(0, 1, "Protein Name")
        col = 2
        for repeat in sorted(repeats):
            worksheet.write(0, col, repeat)
            col += 1
        row = 1
        for entry_id, protein_name, freq in file_data:
            worksheet.write(row, 0, entry_id)
            worksheet.write(row, 1, protein_name)
            col = 2
            for repeat in sorted(repeats):
                worksheet.write(row, col, freq.get(repeat, 0))
                col += 1
            row += 1
    workbook.close()
    output.seek(0)
    return output

# Streamlit UI
st.title("Protein Repeat Analysis with Caching")
analysis_type = st.radio("Select analysis type:", ["Homo", "Hetero", "Both"], index=2)
uploaded_files = st.file_uploader("Upload Excel files", accept_multiple_files=True, type=["xlsx"])

if uploaded_files:
    all_repeats = set()
    all_sequences_data = []
    filenames = []
    for file in uploaded_files:
        excel_data = pd.ExcelFile(file)
        repeats, sequence_data = process_excel(excel_data, analysis_type)
        if repeats is not None:
            all_repeats.update(repeats)
            all_sequences_data.append(sequence_data)
            filenames.append(file.name)
    if all_sequences_data:
        st.toast(f"Processed {len(uploaded_files)} file(s) successfully.")
        excel_file = create_excel(all_sequences_data, all_repeats, filenames)
        st.download_button(
            label="Download Excel file",
            data=excel_file,
            file_name="protein_repeat_results.xlsx",
            mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
        )
        if st.checkbox("Show Results Table"):
            rows = []
            for file_index, file_data in enumerate(all_sequences_data):
                filename = filenames[file_index]
                for entry_id, protein_name, freq in file_data:
                    row = {"Filename": filename, "Entry": entry_id, "Protein Name": protein_name}
                    row.update({repeat: freq.get(repeat, 0) for repeat in sorted(all_repeats)})
                    rows.append(row)
            result_df = pd.DataFrame(rows)
            st.dataframe(result_df)