广电网络公司优秀营销案例广州seo网站多少钱

函数 featurize 的主要作用是将一批蛋白质序列和结构信息转化为深度学习模型可以接受的特征矩阵。它在处理蛋白质多链结构（即多个链的蛋白质复合体）时，考虑了可见链和被掩码链的区分。

代码：

import torch
import numpy as np
import csv
import time
import os
import randomdef featurize(batch, device):alphabet = 'ACDEFGHIKLMNPQRSTVWYX'B = len(batch)lengths = np.array([len(b['seq']) for b in batch], dtype=np.int32) #sum of chain seq lengthsL_max = max([len(b['seq']) for b in batch])X = np.zeros([B, L_max, 4, 3])residue_idx = -100*np.ones([B, L_max], dtype=np.int32) #residue idx with jumps across chainschain_M = np.zeros([B, L_max], dtype=np.int32) #1.0 for the bits that need to be predicted, 0.0 for the bits that are givenmask_self = np.ones([B, L_max, L_max], dtype=np.int32) #for interface loss calculation - 0.0 for self interaction, 1.0 for otherchain_encoding_all = np.zeros([B, L_max], dtype=np.int32) #integer encoding for chains 0, 0, 0,...0, 1, 1,..., 1, 2, 2, 2...S = np.zeros([B, L_max], dtype=np.int32) #sequence AAs integersinit_alphabet = ['A', 'B', 'C', 'D', 'E', 'F', 'G','H', 'I', 'J','K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T','U', 'V','W','X', 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f', 'g','h', 'i', 'j','k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't','u', 'v','w','x', 'y', 'z']extra_alphabet = [str(item) for item in list(np.arange(300))]chain_letters = init_alphabet + extra_alphabetfor i, b in enumerate(batch):masked_chains = b['masked_list']visible_chains = b['visible_list']all_chains = masked_chains + visible_chainsvisible_temp_dict = {}masked_temp_dict = {}for step, letter in enumerate(all_chains):chain_seq = b[f'seq_chain_{letter}']if letter in visible_chains:visible_temp_dict[letter] = chain_seqelif letter in masked_chains:masked_temp_dict[letter] = chain_seqfor km, vm in masked_temp_dict.items():for kv, vv in visible_temp_dict.items():if v