create_diffusion_dataset.py

import pandas as pd
import numpy as np
import networkx as nx
from typing import List, Dict, Tuple, Set, Optional
import json
import random
from collections import defaultdict, Counter
import multiprocessing as mp
from functools import partial
import time
import requests
from urllib.parse import quote
import xml.etree.ElementTree as ET
from time import sleep
import re
import os
import fcntl
import tempfile
import shutil

# Global configuration
MAX_PROTEINS_TO_PROCESS = 1000
TRUNCATE_MIDDLE_ABOVE_THIS = 120

class ProteinDataEnricher:
    """
    Handles fetching and caching protein sequence and structural data with thread-safe persistent cache
    """
    def __init__(self, cache_dir: str = "protein_cache"):
        self.cache_dir = cache_dir
        self.uniprot_cache = {}
        self.alphafold_cache = {}
        self.session = requests.Session()
        self.session.headers.update({'User-Agent': 'ProteinNetworkDataset/1.0'})
        
        # Create cache directory if it doesn't exist
        os.makedirs(cache_dir, exist_ok=True)
        
        # Load existing caches
        self._load_caches()
        
    def _get_cache_files(self):
        """Get cache file paths"""
        uniprot_cache_file = os.path.join(self.cache_dir, "uniprot_cache.json")
        alphafold_cache_file = os.path.join(self.cache_dir, "alphafold_cache.json")
        return uniprot_cache_file, alphafold_cache_file
        
    def _load_caches(self):
        """Load existing caches from disk"""
        uniprot_cache_file, alphafold_cache_file = self._get_cache_files()
        
        # Load UniProt cache
        if os.path.exists(uniprot_cache_file):
            try:
                with open(uniprot_cache_file, 'r') as f:
                    self.uniprot_cache = json.load(f)
                print(f"Loaded UniProt cache with {len(self.uniprot_cache)} entries")
            except Exception as e:
                print(f"Error loading UniProt cache: {e}")
                self.uniprot_cache = {}
        
        # Load AlphaFold cache
        if os.path.exists(alphafold_cache_file):
            try:
                with open(alphafold_cache_file, 'r') as f:
                    self.alphafold_cache = json.load(f)
                print(f"Loaded AlphaFold cache with {len(self.alphafold_cache)} entries")
            except Exception as e:
                print(f"Error loading AlphaFold cache: {e}")
                self.alphafold_cache = {}
    
    def _safe_merge_and_save_cache(self, cache_type: str, new_data: Dict):
        """
        Safely merge new cache data with existing cache using file locking
        """
        if cache_type == "uniprot":
            cache_file = os.path.join(self.cache_dir, "uniprot_cache.json")
        elif cache_type == "alphafold":
            cache_file = os.path.join(self.cache_dir, "alphafold_cache.json")
        else:
            raise ValueError(f"Unknown cache type: {cache_type}")
        
        # Create lock file
        lock_file = cache_file + ".lock"
        
        try:
            # Acquire lock
            with open(lock_file, 'w') as lock:
                fcntl.flock(lock.fileno(), fcntl.LOCK_EX)
                
                # Load current cache from file
                current_cache = {}
                if os.path.exists(cache_file):
                    try:
                        with open(cache_file, 'r') as f:
                            current_cache = json.load(f)
                    except Exception as e:
                        print(f"Error loading {cache_type} cache for merge: {e}")
                        current_cache = {}
                
                # Merge new data with current cache
                current_cache.update(new_data)
                
                # Atomic write: write to temp file then rename
                temp_file = cache_file + ".tmp"
                with open(temp_file, 'w') as f:
                    json.dump(current_cache, f, indent=2)
                
                # Atomic rename
                shutil.move(temp_file, cache_file)
                
                print(f"Merged and saved {cache_type} cache: {len(current_cache)} total entries")
                
        except Exception as e:
            print(f"Error saving {cache_type} cache: {e}")
        finally:
            # Clean up lock file
            try:
                os.remove(lock_file)
            except:
                pass
    
    def get_uniprot_info(self, protein_name: str) -> Optional[Dict]:
        """
        Fetch UniProt information for a protein
        """
        if protein_name in self.uniprot_cache:
            return self.uniprot_cache[protein_name]
        
        try:
            # Search UniProt for the protein
            search_url = f"https://rest.uniprot.org/uniprotkb/search"
            params = {
                'query': f'gene_exact:{protein_name} OR protein_name:{protein_name}',
                'format': 'json',
                'size': 1
            }
            
            response = self.session.get(search_url, params=params, timeout=10)
            if response.status_code == 200:
                data = response.json()
                if data.get('results'):
                    entry = data['results'][0]
                    uniprot_info = {
                        'uniprot_id': entry.get('primaryAccession'),
                        'protein_name': entry.get('proteinDescription', {}).get('recommendedName', {}).get('fullName', {}).get('value'),
                        'gene_names': [gn.get('geneName', {}).get('value') for gn in entry.get('genes', []) if gn.get('geneName')],
                        'organism': entry.get('organism', {}).get('scientificName'),
                        'sequence': entry.get('sequence', {}).get('value'),
                        'sequence_length': entry.get('sequence', {}).get('length'),
                        'function': entry.get('comments', [{}])[0].get('texts', [{}])[0].get('value') if entry.get('comments') else None
                    }
                    self.uniprot_cache[protein_name] = uniprot_info
                    return uniprot_info
                    
        except Exception as e:
            print(f"Error fetching UniProt data for {protein_name}: {e}")
        
        self.uniprot_cache[protein_name] = None
        return None
    
    def get_alphafold_info(self, uniprot_id: str) -> Optional[Dict]:
        """
        Fetch AlphaFold structural information for a UniProt ID
        """
        if not uniprot_id:
            return None
            
        cache_key = uniprot_id
        if cache_key in self.alphafold_cache:
            return self.alphafold_cache[cache_key]
        
        try:
            # Check if AlphaFold structure exists
            af_url = f"https://alphafold.ebi.ac.uk/api/prediction/{uniprot_id}"
            response = self.session.get(af_url, timeout=10)
            
            if response.status_code == 200:
                af_data = response.json()
                if af_data:
                    entry = af_data[0] if isinstance(af_data, list) else af_data
                    
                    alphafold_info = {
                        'alphafold_id': entry.get('entryId'),
                        'model_confidence': entry.get('modelConfidence'),
                        'model_url': entry.get('pdbUrl'),
                        'confidence_score': entry.get('modelConfidence'),
                        'structure_coverage': f"{entry.get('uniprotStart', 1)}-{entry.get('uniprotEnd', 'end')}",
                        'has_structure': True
                    }
                    self.alphafold_cache[cache_key] = alphafold_info
                    return alphafold_info
                    
        except Exception as e:
            print(f"Error fetching AlphaFold data for {uniprot_id}: {e}")
        
        self.alphafold_cache[cache_key] = {'has_structure': False}
        return {'has_structure': False}
    
    def get_protein_enriched_data(self, protein_name: str) -> Dict:
        """
        Get combined UniProt and AlphaFold data for a protein
        """
        enriched_data = {'protein_name': protein_name}
        
        # Get UniProt info
        uniprot_info = self.get_uniprot_info(protein_name)
        if uniprot_info:
            enriched_data.update(uniprot_info)
            
            # Get AlphaFold info if we have a UniProt ID
            if uniprot_info.get('uniprot_id'):
                alphafold_info = self.get_alphafold_info(uniprot_info['uniprot_id'])
                if alphafold_info:
                    enriched_data.update(alphafold_info)
        
        # Add rate limiting
        sleep(0.1)  # Small delay to be respectful to APIs
        return enriched_data
    
    def save_cache(self):
        """Save caches using safe merge strategy"""
        # Only save entries that are not None/empty
        uniprot_to_save = {k: v for k, v in self.uniprot_cache.items() if v is not None}
        alphafold_to_save = {k: v for k, v in self.alphafold_cache.items() if v is not None}
        
        if uniprot_to_save:
            self._safe_merge_and_save_cache("uniprot", uniprot_to_save)
        
        if alphafold_to_save:
            self._safe_merge_and_save_cache("alphafold", alphafold_to_save)

def enrich_proteins_worker(args):
    """
    Worker function to enrich a batch of proteins with sequence/structure data
    """
    protein_names_batch, cache_dir, worker_id = args
    enricher = ProteinDataEnricher(cache_dir)
    enriched_proteins = {}
    
    print(f"Worker {worker_id}: Processing {len(protein_names_batch)} proteins")
    
    for i, protein_name in enumerate(protein_names_batch):
        enriched_proteins[protein_name] = enricher.get_protein_enriched_data(protein_name)
        
        # Save cache every 25 proteins to avoid losing work
        if (i + 1) % 25 == 0:
            enricher.save_cache()
            print(f"Worker {worker_id}: Saved cache after {i + 1} proteins")
    
    # Final save
    enricher.save_cache()
    print(f"Worker {worker_id}: Completed batch, final cache save")
    
    return enriched_proteins

def extract_neighborhood_worker(args):
    """
    Worker function for parallel neighborhood extraction
    """
    center_protein, interactions_by_protein, all_interactions, max_size = args
    
    # Get direct neighbors
    neighbors = set()
    relevant_interactions = []
    
    for interaction in interactions_by_protein[center_protein]:
        other_protein = (interaction['protein_b'] if interaction['protein_a'] == center_protein 
                       else interaction['protein_a'])
        neighbors.add(other_protein)
        relevant_interactions.append(interaction)
    
    # Limit neighborhood size
    if len(neighbors) > max_size - 1:
        neighbors = set(random.sample(list(neighbors), max_size - 1))
    
    # Get all interactions within this neighborhood
    neighborhood_proteins = {center_protein} | neighbors
    neighborhood_interactions = []
    
    for interaction in all_interactions:
        if (interaction['protein_a'] in neighborhood_proteins and 
            interaction['protein_b'] in neighborhood_proteins):
            neighborhood_interactions.append(interaction)
    
    return {
        'center_protein': center_protein,
        'proteins': sorted(list(neighborhood_proteins)),
        'interactions': neighborhood_interactions
    }

def create_conversation_examples_worker(args):
    """
    Worker function for parallel conversation creation
    """
    neighborhood, enriched_proteins = args
    creator = ProteinNetworkConversationDataset("")  # Dummy instance for methods
    creator.enriched_proteins = enriched_proteins  # Set the enriched protein data
    conversations = []
    
    # Task 1: Complete protein network given protein list
    conversations.extend(creator.create_protein_list_to_network_examples(neighborhood))
    
    # Task 2: Predict interactions for new protein
    conversations.extend(creator.create_new_protein_prediction_examples(neighborhood))
    
    # Task 3: Complete partial network
    conversations.extend(creator.create_partial_network_completion_examples(neighborhood))
    
    # Task 4: Network property prediction
    conversations.extend(creator.create_network_property_examples(neighborhood))
    
    return conversations

class ProteinNetworkConversationDataset:
    def __init__(self, filename: str, cache_dir: str = "protein_cache"):
        """
        Create conversational dataset for protein network prediction using diffusion models
        """
        self.filename = filename
        self.cache_dir = cache_dir
        self.df = None
        self.graph = nx.Graph()
        self.protein_to_id = {}
        self.id_to_protein = {}
        self.interactions_by_protein = defaultdict(list)
        self.enriched_proteins = {}  # Cache for enriched protein data
        self.enricher = ProteinDataEnricher(cache_dir)
        
    def load_and_parse_biogrid(self):
        """Load and parse BioGRID data"""
        print("Loading BioGRID data...")
        self.df = pd.read_csv(
            self.filename, 
            sep='\t', 
            comment='#',
            low_memory=False,
            dtype=str
        )
        
        # Based on your sample output - adjust column indices as needed
        protein_a_col = 7  # MAP2K4, MYPN, etc.
        protein_b_col = 8  # FLNC, etc.
        interaction_type_col = 11  # "Two-hybrid", "physical", etc.
        
        interactions = []
        protein_set = set()
        # Use set to track unique interactions and avoid duplicates
        seen_interactions = set()
        
        for idx, row in self.df.iterrows():
            try:
                protein_a = str(row.iloc[protein_a_col]).strip()
                protein_b = str(row.iloc[protein_b_col]).strip()
                interaction_type = str(row.iloc[interaction_type_col]).strip()
                
                if protein_a in ['-', 'nan', ''] or protein_b in ['-', 'nan', '']:
                    continue
                
                # Skip self-interactions
                if protein_a == protein_b:
                    continue
                    
                # Create canonical interaction key (sorted to avoid A-B vs B-A duplicates)
                interaction_key = tuple(sorted([protein_a, protein_b]) + [interaction_type])
                
                # Skip if we've already seen this exact interaction
                if interaction_key in seen_interactions:
                    continue
                    
                seen_interactions.add(interaction_key)
                protein_set.add(protein_a)
                protein_set.add(protein_b)
                
                interaction = {
                    'protein_a': protein_a,
                    'protein_b': protein_b,
                    'interaction_type': interaction_type
                }
                interactions.append(interaction)
                
                # Build protein interaction index
                self.interactions_by_protein[protein_a].append(interaction)
                self.interactions_by_protein[protein_b].append(interaction)
                
            except Exception:
                continue
        
        print(f"Extracted {len(interactions)} valid unique interactions")
        print(f"Found {len(protein_set)} unique proteins")
        
        return interactions, sorted(list(protein_set))
    
    def build_network_neighborhoods(self, interactions, proteins, min_connections=3, max_connections=15):
        """
        Build subnetworks around high-degree proteins for training examples
        """
        # Count protein degrees
        protein_degrees = Counter()
        for interaction in interactions:
            protein_degrees[interaction['protein_a']] += 1
            protein_degrees[interaction['protein_b']] += 1
        
        # Find proteins with moderate connectivity (good for examples)
        candidate_proteins = [
            protein for protein, degree in protein_degrees.items()
            if min_connections <= degree <= max_connections
        ]
        
        print(f"Found {len(candidate_proteins)} proteins with degree {min_connections}-{max_connections}")
        
        # Limit candidate proteins for processing
        limited_proteins = candidate_proteins[:MAX_PROTEINS_TO_PROCESS]
        print(f"Processing {len(limited_proteins)} proteins with multiprocessing...")
        
        # Prepare arguments for parallel processing
        worker_args = [
            (protein, self.interactions_by_protein, interactions, 10) 
            for protein in limited_proteins
        ]
        
        # Use multiprocessing to extract neighborhoods in parallel
        num_processes = min(int(mp.cpu_count()/2), len(limited_proteins))
        print(f"Using {num_processes} processes")
        
        neighborhoods = []
        with mp.Pool(processes=num_processes) as pool:
            results = pool.map(extract_neighborhood_worker, worker_args)
            
            # Filter neighborhoods with minimum viable size
            neighborhoods = [
                neighborhood for neighborhood in results 
                if len(neighborhood['proteins']) >= 3
            ]
        
        return neighborhoods
    
    def enrich_proteins_with_data(self, proteins: List[str]):
        """
        Enrich proteins with UniProt and AlphaFold data using multiprocessing with thread-safe persistent cache
        """
        print(f"Enriching {len(proteins)} proteins with sequence and structural data...")
        
        # Check how many proteins are already cached
        cached_count = 0
        proteins_to_fetch = []
        
        for protein in proteins:
            # Check if protein is already in enricher's cache
            if (protein in self.enricher.uniprot_cache):
                cached_count += 1
                # Use cached data
                self.enriched_proteins[protein] = self.enricher.get_protein_enriched_data(protein)
            else:
                proteins_to_fetch.append(protein)
        
        print(f"Found {cached_count} proteins in cache, need to fetch {len(proteins_to_fetch)} new proteins")
        
        if not proteins_to_fetch:
            print("All proteins found in cache!")
            return self.enriched_proteins
        
        # Split proteins to fetch into batches for parallel processing
        batch_size = 25  # Smaller batch size for better cache management
        protein_batches = [proteins_to_fetch[i:i+batch_size] for i in range(0, len(proteins_to_fetch), batch_size)]
        
        # Prepare arguments with cache directory and worker IDs
        worker_args = [(batch, self.cache_dir, i) for i, batch in enumerate(protein_batches)]
        
        # Use multiprocessing to enrich proteins
        num_processes = min(int(mp.cpu_count()/2), len(protein_batches))
        print(f"Using {num_processes} processes for protein enrichment with thread-safe caching")
        
        enrichment_start = time.time()
        with mp.Pool(processes=num_processes) as pool:
            results = pool.map(enrich_proteins_worker, worker_args)
        
        # Combine results
        for batch_result in results:
            self.enriched_proteins.update(batch_result)
        
        # Reload the main enricher's cache to get all the merged data
        print("Reloading cache to get all merged data...")
        self.enricher._load_caches()
        
        # Fill in any cached data we might have missed
        for protein in proteins:
            if protein not in self.enriched_proteins and protein in self.enricher.uniprot_cache:
                self.enriched_proteins[protein] = self.enricher.get_protein_enriched_data(protein)
        
        enrichment_time = time.time() - enrichment_start
        successful_enrichments = sum(1 for data in self.enriched_proteins.values() 
                                   if data.get('uniprot_id') is not None)
        
        print(f"Protein enrichment completed in {enrichment_time:.2f} seconds")
        print(f"Successfully enriched {successful_enrichments}/{len(proteins)} proteins")
        print(f"Final cache sizes - UniProt: {len(self.enricher.uniprot_cache)}, AlphaFold: {len(self.enricher.alphafold_cache)}")
        
        return self.enriched_proteins
    
    def extract_neighborhood(self, center_protein, interactions, max_size=10):
        """
        Extract neighborhood around a protein
        """
        # Get direct neighbors
        neighbors = set()
        relevant_interactions = []
        
        for interaction in self.interactions_by_protein[center_protein]:
            other_protein = (interaction['protein_b'] if interaction['protein_a'] == center_protein 
                           else interaction['protein_a'])
            neighbors.add(other_protein)
            relevant_interactions.append(interaction)
        
        # Limit neighborhood size
        if len(neighbors) > max_size - 1:
            neighbors = set(random.sample(list(neighbors), max_size - 1))
        
        # Get all interactions within this neighborhood
        neighborhood_proteins = {center_protein} | neighbors
        neighborhood_interactions = []
        
        for interaction in interactions:
            if (interaction['protein_a'] in neighborhood_proteins and 
                interaction['protein_b'] in neighborhood_proteins):
                neighborhood_interactions.append(interaction)
        
        return {
            'center_protein': center_protein,
            'proteins': sorted(list(neighborhood_proteins)),
            'interactions': neighborhood_interactions
        }
    
    def create_conversation_examples(self, neighborhoods):
        """
        Create different types of conversation examples for diffusion training (parallelized)
        """
        print(f"Creating conversation examples for {len(neighborhoods)} neighborhoods using multiprocessing...")
        
        # Prepare arguments for parallel processing (include enriched protein data)
        worker_args = [(neighborhood, self.enriched_proteins) for neighborhood in neighborhoods]
        
        # Use multiprocessing to create conversation examples in parallel
        num_processes = min(int(mp.cpu_count()/2), len(neighborhoods))
        print(f"Using {num_processes} processes for conversation creation")
        
        conversations = []
        with mp.Pool(processes=num_processes) as pool:
            results = pool.map(create_conversation_examples_worker, worker_args)
            
            # Flatten the results
            for result in results:
                conversations.extend(result)
        
        return conversations
    
    def create_protein_list_to_network_examples(self, neighborhood):
        """
        Context: List of proteins with sequence/structure info
        Generation: Complete interaction network
        """
        examples = []
        proteins = neighborhood['proteins']
        interactions = neighborhood['interactions']
        
        # Create enriched network representation
        network_text = self.format_enriched_network_as_text(proteins, interactions)
        
        system_msg = {
            "role": "system",
            "content": "You are a protein interaction prediction system. Given a list of proteins with their sequence and structural information, predict all likely interactions between them based on biological knowledge, sequence similarity, and structural compatibility."
        }
        
        # Create enriched protein context
        protein_context = self.format_proteins_with_context(proteins)
        
        user_msg = {
            "role": "user", 
            "content": f"Predict the protein interaction network for these proteins:\n\n{protein_context}"
        }
        
        assistant_msg = {
            "role": "assistant",
            "content": network_text
        }
        
        conversation = [system_msg, user_msg, assistant_msg]
        examples.append({"updated": conversation})
        
        return examples
    
    def create_new_protein_prediction_examples(self, neighborhood):
        """
        Context: Known network + new protein with sequence/structure info
        Generation: Interactions for the new protein
        """
        examples = []
        if len(neighborhood['proteins']) < 4:
            return examples
            
        proteins = neighborhood['proteins']
        interactions = neighborhood['interactions']
        
        # Remove one protein and its interactions for prediction
        target_protein = random.choice(proteins)
        remaining_proteins = [p for p in proteins if p != target_protein]
        
        # Known network (without target protein)
        known_interactions = [
            i for i in interactions 
            if target_protein not in [i['protein_a'], i['protein_b']]
        ]
        
        # Target interactions (what we want to predict)
        target_interactions = [
            i for i in interactions
            if target_protein in [i['protein_a'], i['protein_b']]
        ]
        
        if not target_interactions:
            return examples
        
        known_network_text = self.format_enriched_network_as_text(remaining_proteins, known_interactions)
        target_network_text = self.format_interactions_as_text(target_interactions)
        
        # Get enriched info for target protein
        target_enriched_data = self.enriched_proteins.get(target_protein, {})
        target_context = self.format_proteins_with_context([target_protein])
        
        system_msg = {
            "role": "system",
            "content": "You are a protein interaction prediction system. Given a known protein network and a new protein with sequence and structural information, predict which proteins in the network the new protein will interact with based on sequence similarity, structural compatibility, and functional relationships."
        }
        
        user_msg = {
            "role": "user",
            "content": f"Known protein network:\n{known_network_text}\n\nNew protein to integrate:\n{target_context}\n\nPredict the interactions for {target_protein} based on its sequence, structure, and function:"
        }
        
        assistant_msg = {
            "role": "assistant", 
            "content": target_network_text
        }
        
        conversation = [system_msg, user_msg, assistant_msg]
        examples.append({"updated": conversation})
        
        return examples
    
    def create_partial_network_completion_examples(self, neighborhood):
        """
        Context: Partial network with some missing interactions
        Generation: Complete network
        """
        examples = []
        proteins = neighborhood['proteins']
        interactions = neighborhood['interactions']
        
        if len(interactions) < 3:
            return examples
        
        # Randomly hide some interactions
        n_hidden = max(1, len(interactions) // 3)
        hidden_interactions = random.sample(interactions, n_hidden)
        visible_interactions = [i for i in interactions if i not in hidden_interactions]
        
        partial_network_text = self.format_network_as_text(proteins, visible_interactions)
        complete_network_text = self.format_network_as_text(proteins, interactions)
        
        system_msg = {
            "role": "system",
            "content": "You are a protein interaction prediction system. Given a partial protein network, predict the complete network including missing interactions."
        }
        
        user_msg = {
            "role": "user",
            "content": f"Complete this partial protein network:\n{partial_network_text}"
        }
        
        assistant_msg = {
            "role": "assistant",
            "content": complete_network_text
        }
        
        conversation = [system_msg, user_msg, assistant_msg]
        examples.append({"updated": conversation})
        
        return examples
    
    def create_network_property_examples(self, neighborhood):
        """
        Context: Network properties and constraints
        Generation: Network that satisfies those properties
        """
        examples = []
        proteins = neighborhood['proteins']
        interactions = neighborhood['interactions']
        
        # Calculate network properties
        n_proteins = len(proteins)
        n_interactions = len(interactions)
        density = (2 * n_interactions) / (n_proteins * (n_proteins - 1)) if n_proteins > 1 else 0
        
        # Find hub proteins (high degree)
        protein_degrees = Counter()
        for interaction in interactions:
            protein_degrees[interaction['protein_a']] += 1
            protein_degrees[interaction['protein_b']] += 1
        
        hub_proteins = [p for p, degree in protein_degrees.most_common(2)]
        
        network_text = self.format_network_as_text(proteins, interactions)
        
        system_msg = {
            "role": "system",
            "content": "You are a protein interaction network generator. Given network constraints and properties, generate a biologically plausible protein network."
        }
        
        properties_text = (f"Generate a protein network with the following properties:\n"
                         f"- Proteins: {', '.join(proteins)}\n"
                         f"- Network density: approximately {density:.2f}\n"
                         f"- Hub proteins (highly connected): {', '.join(hub_proteins)}\n"
                         f"- Total interactions: approximately {n_interactions}")
        
        user_msg = {
            "role": "user",
            "content": properties_text
        }
        
        assistant_msg = {
            "role": "assistant",
            "content": network_text
        }
        
        conversation = [system_msg, user_msg, assistant_msg]
        examples.append({"updated": conversation})
        
        return examples
    
    def format_network_as_text(self, proteins, interactions):
        """
        Format network as structured text for the model to predict
        """
        # Sort for consistency
        proteins = sorted(proteins)
        
        # Group interactions by type and use sets to avoid duplicates
        interactions_by_type = defaultdict(set)
        for interaction in interactions:
            # Skip self-interactions
            if interaction['protein_a'] == interaction['protein_b']:
                continue
                
            int_type = interaction.get('interaction_type', 'physical')
            # Ensure consistent ordering
            p1, p2 = sorted([interaction['protein_a'], interaction['protein_b']])
            interactions_by_type[int_type].add(f"{p1}--{p2}")
        
        result = f"PROTEINS: {', '.join(proteins)}\n\n"
        
        for int_type, edges in interactions_by_type.items():
            if edges:
                result += f"{int_type.upper()} INTERACTIONS:\n"
                for edge in sorted(edges):  # Sort the set for consistent output
                    result += f"  {edge}\n"
                result += "\n"
        
        total_interactions = sum(len(edges) for edges in interactions_by_type.values())
        result += f"NETWORK SUMMARY: {len(proteins)} proteins, {total_interactions} unique interactions"
        return result.strip()
    
    def format_proteins_with_context(self, proteins: List[str]) -> str:
        """
        Format proteins with their enriched sequence and structural context
        """
        protein_contexts = []
        
        for protein in sorted(proteins):
            enriched_data = self.enriched_proteins.get(protein, {})
            
            context_parts = [f"PROTEIN: {protein}"]
            
            # Add UniProt information
            if enriched_data.get('uniprot_id'):
                context_parts.append(f"  UniProt ID: {enriched_data['uniprot_id']}")
            
            if enriched_data.get('protein_name'):
                context_parts.append(f"  Full Name: {enriched_data['protein_name']}")
            
            if enriched_data.get('organism'):
                context_parts.append(f"  Organism: {enriched_data['organism']}")
            
            # Add sequence information
            if enriched_data.get('sequence_length'):
                context_parts.append(f"  Sequence Length: {enriched_data['sequence_length']} amino acids")
            
            if enriched_data.get('sequence'):
                # Show first 50 and last 20 amino acids
                seq = enriched_data['sequence']
                if len(seq) > TRUNCATE_MIDDLE_ABOVE_THIS:
                    seq_preview = f"{seq[:int(TRUNCATE_MIDDLE_ABOVE_THIS*0.5)]}...{seq[-int(TRUNCATE_MIDDLE_ABOVE_THIS*0.2):]}"
                else:
                    seq_preview = seq
                context_parts.append(f"  Sequence: {seq_preview}")
            
            # Add structural information
            if enriched_data.get('has_structure'):
                context_parts.append(f"  AlphaFold Structure: Available")
                if enriched_data.get('model_confidence'):
                    context_parts.append(f"  Structure Confidence: {enriched_data['model_confidence']}")
                if enriched_data.get('structure_coverage'):
                    context_parts.append(f"  Structure Coverage: residues {enriched_data['structure_coverage']}")
            else:
                context_parts.append(f"  AlphaFold Structure: Not available")
            
            # Add functional information
            if enriched_data.get('function'):
                func_text = enriched_data['function'][:200] + "..." if len(enriched_data['function']) > 200 else enriched_data['function']
                context_parts.append(f"  Function: {func_text}")
            
            protein_contexts.append("\n".join(context_parts))
        
        return "\n\n".join(protein_contexts)
    
    def format_enriched_network_as_text(self, proteins, interactions):
        """
        Format network with enriched protein information
        """
        # First show the basic network
        basic_network = self.format_network_as_text(proteins, interactions)
        
        # Add enriched protein summaries
        enriched_summary = "\n\nPROTEIN DETAILS:\n"
        
        for protein in sorted(proteins):
            enriched_data = self.enriched_proteins.get(protein, {})
            details = [f"{protein}"]
            
            if enriched_data.get('sequence_length'):
                details.append(f"{enriched_data['sequence_length']}aa")
            
            if enriched_data.get('has_structure'):
                confidence = enriched_data.get('model_confidence', 'unknown')
                details.append(f"AlphaFold({confidence})")
            
            if enriched_data.get('organism'):
                org = enriched_data['organism'].split()[0] if ' ' in enriched_data['organism'] else enriched_data['organism']
                details.append(f"{org}")
            
            enriched_summary += f"  {' | '.join(details)}\n"
        
        return basic_network + enriched_summary
    
    def format_interactions_as_text(self, interactions):
        """
        Format just interactions as text
        """
        if not interactions:
            return "No interactions predicted."
        
        # Group interactions by type and use sets to avoid duplicates
        interactions_by_type = defaultdict(set)
        for interaction in interactions:
            # Skip self-interactions
            if interaction['protein_a'] == interaction['protein_b']:
                continue
                
            int_type = interaction.get('interaction_type', 'physical')
            p1, p2 = sorted([interaction['protein_a'], interaction['protein_b']])
            interactions_by_type[int_type].add(f"{p1}--{p2}")
        
        result = ""
        for int_type, edges in interactions_by_type.items():
            if edges:
                result += f"{int_type.upper()} INTERACTIONS:\n"
                for edge in sorted(edges):  # Sort the set for consistent output
                    result += f"  {edge}\n"
                result += "\n"
        
        return result.strip()
    
    def save_conversation_dataset(self, output_file="processed_dataset.json"):
        """
        Create and save the full conversation dataset with enriched protein data
        """
        start_time = time.time()
        
        # Load and process data
        print("Step 1: Loading and parsing data...")
        load_start = time.time()
        interactions, proteins = self.load_and_parse_biogrid()
        load_time = time.time() - load_start
        print(f"Data loading completed in {load_time:.2f} seconds")
        
        # Build neighborhoods
        print("Step 2: Building protein neighborhoods...")
        neighborhood_start = time.time()
        neighborhoods = self.build_network_neighborhoods(interactions, proteins)
        neighborhood_time = time.time() - neighborhood_start
        print(f"Built {len(neighborhoods)} protein neighborhoods in {neighborhood_time:.2f} seconds")
        
        # Enrich proteins with sequence and structural data
        print("Step 3: Enriching proteins with sequence and structural data...")
        enrichment_start = time.time()
        # Get unique proteins from neighborhoods
        unique_proteins = set()
        for neighborhood in neighborhoods:
            unique_proteins.update(neighborhood['proteins'])
        
        self.enrich_proteins_with_data(list(unique_proteins))
        enrichment_time = time.time() - enrichment_start
        print(f"Protein enrichment completed in {enrichment_time:.2f} seconds")
        
        # Create conversation examples
        print("Step 4: Creating conversation examples...")
        conversation_start = time.time()
        conversations = self.create_conversation_examples(neighborhoods)
        conversation_time = time.time() - conversation_start
        print(f"Created {len(conversations)} conversation examples in {conversation_time:.2f} seconds")
        
        # Shuffle the dataset
        print("Step 5: Shuffling and saving dataset...")
        random.shuffle(conversations)
        
        # Save dataset
        with open(output_file, 'w') as f:
            json.dump(conversations, f, indent=2)
        
        # Save enriched protein data separately for reference
        enriched_data_file = output_file.replace('.json', '_protein_data.json')
        with open(enriched_data_file, 'w') as f:
            json.dump(self.enriched_proteins, f, indent=2)
        
        total_time = time.time() - start_time
        print(f"Saved dataset to {output_file}")
        print(f"Saved enriched protein data to {enriched_data_file}")
        print(f"Total processing time: {total_time:.2f} seconds")
        
        # Show examples
        print("\n=== Example Conversations ===")
        for i, conv in enumerate(conversations[:2]):
            print(f"\n--- Example {i+1} ---")
            for msg in conv["updated"]:
                print(f"{msg['role'].upper()}: {msg['content'][:300]}...")
        
        return conversations

# Usage
if __name__ == "__main__":
    # Set random seed for reproducibility
    random.seed(42)
    np.random.seed(42)
    
    print(f"Configuration: Processing up to {MAX_PROTEINS_TO_PROCESS} proteins")
    print(f"Available CPU cores: {int(mp.cpu_count()/2)}")
    print(f"Cache directory: protein_cache/")
    
    creator = ProteinNetworkConversationDataset(
        "./unzipped/BIOGRID-ALL-4.4.246.tab3/BIOGRID-ALL-4.4.246.tab3.txt",
        cache_dir="protein_cache"
    )
    
    conversations = creator.save_conversation_dataset("processed_dataset.json")
    
    print(f"\n=== Dataset Summary ===")
    print(f"Total conversations: {len(conversations)}")
    
    # Count conversation types by system message content
    task_types = Counter()
    for conv in conversations:
        system_content = conv["updated"][0]["content"]
        if "list of proteins" in system_content:
            task_types["protein_list_to_network"] += 1
        elif "new protein" in system_content:
            task_types["new_protein_integration"] += 1
        elif "partial" in system_content:
            task_types["partial_completion"] += 1
        elif "properties" in system_content:
            task_types["property_based_generation"] += 1
    
    print("\nTask distribution:")
    for task, count in task_types.items():
        print(f"  {task}: {count}")
    
    # Print enrichment statistics
    print(f"\n=== Protein Enrichment Summary ===")
    total_proteins = len(creator.enriched_proteins)
    proteins_with_uniprot = sum(1 for data in creator.enriched_proteins.values() 
                               if data.get('uniprot_id') is not None)
    proteins_with_sequence = sum(1 for data in creator.enriched_proteins.values() 
                                if data.get('sequence') is not None)
    proteins_with_structure = sum(1 for data in creator.enriched_proteins.values() 
                                 if data.get('has_structure') == True)
    
    print(f"Total proteins processed: {total_proteins}")
    print(f"Proteins with UniProt data: {proteins_with_uniprot} ({proteins_with_uniprot/total_proteins*100:.1f}%)")
    print(f"Proteins with sequences: {proteins_with_sequence} ({proteins_with_sequence/total_proteins*100:.1f}%)")
    print(f"Proteins with AlphaFold structures: {proteins_with_structure} ({proteins_with_structure/total_proteins*100:.1f}%)")
    
    # Print cache statistics
    print(f"\n=== Cache Statistics ===")
    print(f"UniProt cache entries: {len(creator.enricher.uniprot_cache)}")
    print(f"AlphaFold cache entries: {len(creator.enricher.alphafold_cache)}")
    print(f"Cache files location: {creator.cache_dir}/")