data_utils.py

import os
import numpy as np
import random
import torch
from torch.utils.data import Dataset, DataLoader

class Hg38VariableDataset(Dataset):
    def __init__(self, data_dir='data', window_size=10_000):
        self.data_dir = data_dir
        self.window_size = window_size
        self.chromosomes = self._load_chromosomes()
        self.chromosome_lengths = {k: v.shape[-1] for k, v in self.chromosomes.items()}
        self.chromosome_names = list(self.chromosomes.keys())
        self.nucleotides = np.array(['A', 'T', 'G', 'C', 'N'])
        
    def _load_chromosomes(self):
        chromosomes = {}
        for filename in sorted(os.listdir(self.data_dir)):
            if filename.endswith(".npz"):
                chrom_name = filename.split(".")[0]  # Use file name (e.g., "chr1")
                matrix = np.load(os.path.join(self.data_dir, filename))['data']
                chromosomes[chrom_name] = matrix
        return chromosomes

    def select_nucleotide_fast(self, matrix):
        choices = np.argmax(matrix * np.random.rand(*matrix.shape), axis=0)
        return ''.join(self.nucleotides[choices])

    def __len__(self):
        return sum(self.chromosome_lengths.values()) 

    def __getitem__(self, idx):
        chrom = random.choice(self.chromosome_names)
        chrom_length = self.chromosome_lengths[chrom]
        start = np.random.randint(0, chrom_length - self.window_size)
        sequence = self.select_nucleotide_fast(self.chromosomes[chrom][:, start:start + self.window_size])

        return {
            'chromosome': chrom,
            'start': start,
            'sequence': sequence
        }

# # # Example usage
# dataset = Hg38VariableDataset()
# data_loader = DataLoader(dataset, batch_size=32, shuffle=True)

# # Iterate over batches
# for batch in data_loader:
#     break