バイオインフォマティクス実践ガイド

import scanpy as sc
import pandas as pd
import numpy as np

class SingleCellProcessor:
    """scRNA-seq前処理パイプライン"""
    
    def __init__(self, min_genes=200, min_cells=3):
        self.min_genes = min_genes
        self.min_cells = min_cells
        
    def quality_control(self, adata):
        """
        品質管理とフィルタリング
        
        Args:
            adata: AnnData object
        """
        # ミトコンドリア遺伝子の割合計算
        mito_genes = adata.var_names.str.startswith('MT-')
        adata.obs['percent_mito'] = np.sum(
            adata[:, mito_genes].X, axis=1).A1 / np.sum(adata.X, axis=1).A1
        
        # 品質指標の計算
        adata.obs['n_genes'] = (adata.X > 0).sum(axis=1)
        
        # フィルタリング
        sc.pp.filter_cells(adata, min_genes=self.min_genes)
        sc.pp.filter_genes(adata, min_cells=self.min_cells)
        
        # 外れ値除去
        adata = adata[adata.obs['percent_mito'] < 0.2]
        adata = adata[adata.obs['n_genes'] < 2500]
        
        return adata

def normalize_and_select_features(adata):
    """
    データ正規化と高変動遺伝子の選択
    """
    # Library size normalization
    sc.pp.normalize_total(adata, target_sum=1e4)
    
    # Log transformation
    sc.pp.log1p(adata)
    
    # Highly variable genes selection
    sc.pp.highly_variable_genes(
        adata, min_mean=0.0125, max_mean=3, min_disp=0.5
    )
    
    # Keep only HVGs
    adata = adata[:, adata.var.highly_variable]
    
    return adata

class DimensionReduction:
    """次元削減手法の実装"""
    
    def perform_pca(self, adata, n_comps=50):
        """主成分分析"""
        sc.pp.scale(adata, max_value=10)
        sc.tl.pca(adata, n_comps=n_comps)
        return adata
    
    def perform_umap(self, adata, n_neighbors=30):
        """UMAP埋め込み"""
        sc.pp.neighbors(adata, n_neighbors=n_neighbors)
        sc.tl.umap(adata)
        return adata

def cluster_cells(adata, resolution=0.5):
    """
    Louvainクラスタリング
    """
    sc.tl.louvain(adata, resolution=resolution)
    return adata

class TrajectoryInference:
    """細胞分化軌跡の推定"""
    
    def perform_diffusion_map(self, adata):
        """拡散マップによる軌跡推定"""
        sc.tl.diffmap(adata)
        return adata
    
    def pseudotime_ordering(self, adata, root_cell):
        """擬時間順序付け"""
        # DPT (Diffusion Pseudotime)
        adata.uns['iroot'] = root_cell
        sc.tl.dpt(adata)
        return adata

class SpatialTranscriptomics:
    """空間発現解析"""
    
    def __init__(self, adata, spatial_coords):
        self.adata = adata
        self.spatial_coords = spatial_coords
        
    def identify_spatial_patterns(self):
        """空間的発現パターンの同定"""
        # Moran's I統計量による空間相関
        spatial_genes = []
        
        for gene in self.adata.var_names:
            morans_i = self.calculate_morans_i(
                self.adata[:, gene].X,
                self.spatial_coords
            )
            if morans_i > 0.3:  # 閾値
                spatial_genes.append(gene)
                
        return spatial_genes