File size: 8,979 Bytes
4d12519 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 | import json
from pytorch_lightning import LightningDataModule
from torch.utils.data import DataLoader, ConcatDataset, Dataset
from data_provider.stage1_dm import SwissProtDataset, OntoProteinDataset
import pandas as pd
class GO_BP(Dataset):
def __init__(self, data_path, prompt='', return_prompt=False):
super(GO_BP, self).__init__()
self.data_path = data_path
self.user_prompt = prompt
self.return_prompt = return_prompt
self.data_list = self._load_and_preprocess(self.data_path)
self.text2id = self._build_text_vocab()
def _load_and_preprocess(self, data_path):
data_list = []
df = pd.read_csv(data_path)
for _, row in df.iterrows():
try:
prot_seq = str(row['question']).strip()
result = str(row['answer']).strip()
text_seq = f"<answer>{result}</answer>\n"
prompt = f"""
【Task】Predict the biological processes involving the given protein.
【Background】Each process is represented by a GO-BP term (e.g., GO:0008150), describing a series of molecular events relevant to protein function.
【Output Format】List the predicted GO-BP terms, separated by commas, and wrap them in <answer> </answer> tags.
Example: <answer>GO:0008150, GO:0009987, GO:0050896</answer>
"""
if self.user_prompt:
prompt += self.user_prompt
# extra可以返回原始feather字符串，也可以返回feather_vals
# 或 feather_raw
data_list.append((prot_seq, text_seq, prompt))
except Exception as e:
print(f"警告: 跳过有问题的行: {row}，原因: {e}")
return data_list
def _build_text_vocab(self):
text2id = {}
for _, text_seq, _ in self.data_list:
if text_seq not in text2id:
text2id[text_seq] = len(text2id)
return text2id
def shuffle(self):
random.shuffle(self.data_list)
return self
def __len__(self):
return len(self.data_list)
def __getitem__(self, index):
prot_seq, text_seq, prompt = self.data_list[index]
if self.return_prompt:
return prot_seq, prompt, text_seq,index
return prot_seq, text_seq, index
class GO_CC(Dataset):
def __init__(self, data_path, prompt='', return_prompt=False):
super(GO_CC, self).__init__()
self.data_path = data_path
self.user_prompt = prompt
self.return_prompt = return_prompt
self.data_list = self._load_and_preprocess(self.data_path)
self.text2id = self._build_text_vocab()
def _load_and_preprocess(self, data_path):
data_list = []
df = pd.read_csv(data_path)
for _, row in df.iterrows():
try:
prot_seq = str(row['question']).strip()
result = str(row['answer']).strip()
text_seq = f"<answer>{result}</answer>\n"
prompt = f"""
【Task】Predict the cellular components associated with this protein.
【Background】Each location is represented by a GO-CC term (e.g., GO:0005737), indicating where the protein functions within the cell.
【Output Format】List the predicted GO-CC terms, separated by commas, and wrap them in <answer> </answer> tags.
Example: <answer>GO:0005737, GO:0005829, GO:0005886</answer>
"""
if self.user_prompt:
prompt += self.user_prompt
# extra可以返回原始feather字符串，也可以返回feather_vals
# 或 feather_raw
data_list.append((prot_seq, text_seq, prompt))
except Exception as e:
print(f"警告: 跳过有问题的行: {row}，原因: {e}")
return data_list
def _build_text_vocab(self):
text2id = {}
for _, text_seq, _ in self.data_list:
if text_seq not in text2id:
text2id[text_seq] = len(text2id)
return text2id
def shuffle(self):
random.shuffle(self.data_list)
return self
def __len__(self):
return len(self.data_list)
def __getitem__(self, index):
prot_seq, text_seq, prompt = self.data_list[index]
if self.return_prompt:
return prot_seq, prompt, text_seq,index
return prot_seq, text_seq, index
class GO_MF(Dataset):
def __init__(self, data_path, prompt='', return_prompt=False):
super(GO_MF, self).__init__()
self.data_path = data_path
self.user_prompt = prompt
self.return_prompt = return_prompt
self.data_list = self._load_and_preprocess(self.data_path)
self.text2id = self._build_text_vocab()
def _load_and_preprocess(self, data_path):
data_list = []
df = pd.read_csv(data_path)
for _, row in df.iterrows():
try:
prot_seq = str(row['question']).strip()
result = str(row['answer']).strip()
text_seq = f"<answer>{result}</answer>\n"
prompt = f"""
【Task】Predict the molecular functions performed by this protein.
【Background】Each function is represented by a GO-MF term (e.g., GO:0003677), describing specific biochemical activities of the protein.
【Output Format】List the predicted GO-MF terms, separated by commas, and wrap them in <answer> </answer> tags.
Example: <answer>GO:0003677, GO:0005524, GO:0016787</answer>
"""
if self.user_prompt:
prompt += self.user_prompt
# extra可以返回原始feather字符串，也可以返回feather_vals
# 或 feather_raw
data_list.append((prot_seq, text_seq, prompt))
except Exception as e:
print(f"警告: 跳过有问题的行: {row}，原因: {e}")
return data_list
def _build_text_vocab(self):
text2id = {}
for _, text_seq, _ in self.data_list:
if text_seq not in text2id:
text2id[text_seq] = len(text2id)
return text2id
def shuffle(self):
random.shuffle(self.data_list)
return self
def __len__(self):
return len(self.data_list)
def __getitem__(self, index):
prot_seq, text_seq, prompt = self.data_list[index]
if self.return_prompt:
return prot_seq, prompt, text_seq,index
return prot_seq, text_seq, index
class EC(Dataset):
def __init__(self, data_path, prompt='', return_prompt=False):
super(EC, self).__init__()
self.data_path = data_path
self.user_prompt = prompt
self.return_prompt = return_prompt
self.data_list = self._load_and_preprocess(self.data_path)
self.text2id = self._build_text_vocab()
def _load_and_preprocess(self, data_path):
data_list = []
df = pd.read_csv(data_path)
for _, row in df.iterrows():
try:
name = str(row['name']).strip()
# 先按“-”分割，得到结构信息和 UniProt ID
structure_part, uniprot_id = name.split('-') # '3r7t_A', 'Q9PMG4'
# 再按“_”分割结构信息，得到 PDB ID 和链ID
pdb_id, chain_id = structure_part.split('_') # '3r7t', 'A'
prot_seq = str(row['aa_seq']).strip()
result = str(row['label']).strip()
text_seq = f"<answer>{result}</answer>\n"
prompt = f"""
The information provided above is protein information, one of the chains of its crystal structure {pdb_id}, named {chain_id}, and numbered {uniprot_id} in the Uniprot sequence database.
Based on this information, the possible enzyme activity is inferred and the corresponding EC number is predicted.
【Output Format】List predicted EC numbers, separated by commas, wrapped in <answer> </answer> tags.
"""
if self.user_prompt:
prompt += self.user_prompt
# extra可以返回原始feather字符串，也可以返回feather_vals
# 或 feather_raw
data_list.append((prot_seq, text_seq, prompt))
except Exception as e:
print(f"警告: 跳过有问题的行: {row}，原因: {e}")
return data_list
def _build_text_vocab(self):
text2id = {}
for _, text_seq, _ in self.data_list:
if text_seq not in text2id:
text2id[text_seq] = len(text2id)
return text2id
def shuffle(self):
random.shuffle(self.data_list)
return self
def __len__(self):
return len(self.data_list)
def __getitem__(self, index):
prot_seq, text_seq, prompt = self.data_list[index]
if self.return_prompt:
return prot_seq, prompt, text_seq,index
return prot_seq, text_seq, index |