DiffLinker / app.py
igashov's picture
updated code
88b37fb
raw
history blame
23.4 kB
import argparse
import shutil
import gradio as gr
import numpy as np
import os
import torch
import output
from rdkit import Chem
from src import const
from src.datasets import (
get_dataloader, collate_with_fragment_edges,
collate_with_fragment_without_pocket_edges,
parse_molecule, MOADDataset
)
from src.lightning import DDPM
from src.linker_size_lightning import SizeClassifier
from src.generation import generate_linkers, try_to_convert_to_sdf, get_pocket
from zipfile import ZipFile
MIN_N_STEPS = 100
MAX_N_STEPS = 500
MAX_BATCH_SIZE = 20
MODELS_METADATA = {
'geom_difflinker': {
'link': 'https://zenodo.org/record/7121300/files/geom_difflinker.ckpt?download=1',
'path': 'models/geom_difflinker.ckpt',
},
'geom_difflinker_given_anchors': {
'link': 'https://zenodo.org/record/7775568/files/geom_difflinker_given_anchors.ckpt?download=1',
'path': 'models/geom_difflinker_given_anchors.ckpt',
},
'pockets_difflinker': {
# 'link': 'https://zenodo.org/record/7775568/files/pockets_difflinker_full_no_anchors.ckpt?download=1',
# 'path': 'models/pockets_difflinker.ckpt',
'link': 'https://zenodo.org/records/10988017/files/pockets_difflinker_full_no_anchors_fc_pdb_excluded.ckpt?download=1',
'path': 'models/pockets_difflinker_full_no_anchors_fc_pdb_excluded.ckpt',
},
'pockets_difflinker_given_anchors': {
# 'link': 'https://zenodo.org/record/7775568/files/pockets_difflinker_full.ckpt?download=1',
# 'path': 'models/pockets_difflinker_given_anchors.ckpt',
'link': 'https://zenodo.org/records/10988017/files/pockets_difflinker_full_fc_pdb_excluded.ckpt?download=1',
'path': 'models/pockets_difflinker_full_fc_pdb_excluded.ckpt',
},
}
parser = argparse.ArgumentParser()
parser.add_argument('--ip', type=str, default=None)
args = parser.parse_args()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f'Device: {device}')
os.makedirs("results", exist_ok=True)
size_gnn_path = 'models/geom_size_gnn.ckpt'
size_nn = SizeClassifier.load_from_checkpoint('models/geom_size_gnn.ckpt', map_location=device).eval().to(device)
print('Loaded SizeGNN model')
diffusion_models = {}
for model_name, metadata in MODELS_METADATA.items():
diffusion_path = metadata['path']
diffusion_models[model_name] = DDPM.load_from_checkpoint(diffusion_path, map_location=device).eval().to(device)
print(f'Loaded model {model_name}')
print(os.curdir)
print(os.path.abspath(os.curdir))
print(os.listdir(os.curdir))
def read_molecule_content(path):
with open(path, "r") as f:
return "".join(f.readlines())
def read_molecule(path):
if path.endswith('.pdb'):
return Chem.MolFromPDBFile(path, sanitize=False, removeHs=True)
elif path.endswith('.mol'):
return Chem.MolFromMolFile(path, sanitize=False, removeHs=True)
elif path.endswith('.mol2'):
return Chem.MolFromMol2File(path, sanitize=False, removeHs=True)
elif path.endswith('.sdf'):
return Chem.SDMolSupplier(path, sanitize=False, removeHs=True)[0]
raise Exception('Unknown file extension')
def read_molecule_file(in_file, allowed_extentions):
if isinstance(in_file, str):
path = in_file
else:
path = in_file.name
extension = path.split('.')[-1]
if extension not in allowed_extentions:
msg = output.INVALID_FORMAT_MSG.format(extension=extension)
return None, None, msg
try:
mol = read_molecule(path)
except Exception as e:
e = str(e).replace('\'', '')
msg = output.ERROR_FORMAT_MSG.format(message=e)
return None, None, msg
if extension == 'pdb':
content = Chem.MolToPDBBlock(mol)
elif extension in ['mol', 'mol2', 'sdf']:
content = Chem.MolToMolBlock(mol, kekulize=False)
extension = 'mol'
else:
raise NotImplementedError
return content, extension, None
def show_input(in_fragments, in_protein):
vis = ''
if in_fragments is not None and in_protein is None:
vis = show_fragments(in_fragments)
elif in_fragments is None and in_protein is not None:
vis = show_target(in_protein)
elif in_fragments is not None and in_protein is not None:
vis = show_fragments_and_target(in_fragments, in_protein)
return [vis, gr.Dropdown.update(choices=[], value=None, visible=False), None]
def show_fragments(in_fragments):
molecule, extension, html = read_molecule_file(in_fragments, allowed_extentions=['sdf', 'pdb', 'mol', 'mol2'])
if molecule is not None:
html = output.FRAGMENTS_RENDERING_TEMPLATE.format(molecule=molecule, fmt=extension)
return output.IFRAME_TEMPLATE.format(html=html)
def show_target(in_protein):
molecule, extension, html = read_molecule_file(in_protein, allowed_extentions=['pdb'])
if molecule is not None:
html = output.TARGET_RENDERING_TEMPLATE.format(molecule=molecule, fmt=extension)
return output.IFRAME_TEMPLATE.format(html=html)
def show_fragments_and_target(in_fragments, in_protein):
fragments_molecule, fragments_extension, msg = read_molecule_file(in_fragments, ['sdf', 'pdb', 'mol', 'mol2'])
if fragments_molecule is None:
return output.IFRAME_TEMPLATE.format(html=msg)
target_molecule, target_extension, msg = read_molecule_file(in_protein, allowed_extentions=['pdb'])
if fragments_molecule is None:
return output.IFRAME_TEMPLATE.format(html=msg)
html = output.FRAGMENTS_AND_TARGET_RENDERING_TEMPLATE.format(
molecule=fragments_molecule,
fmt=fragments_extension,
target=target_molecule,
target_fmt=target_extension,
)
return output.IFRAME_TEMPLATE.format(html=html)
def clear_fragments_input(in_protein):
vis = ''
if in_protein is not None:
vis = show_target(in_protein)
return [None, vis, gr.Dropdown.update(choices=[], value=None, visible=False), None]
def clear_protein_input(in_fragments):
vis = ''
if in_fragments is not None:
vis = show_fragments(in_fragments)
return [None, vis, gr.Dropdown.update(choices=[], value=None, visible=False), None]
def click_on_example(example):
fragment_fname, target_fname = example
fragment_path = f'examples/{fragment_fname}' if fragment_fname != '' else None
target_path = f'examples/{target_fname}' if target_fname != '' else None
return [fragment_path, target_path] + show_input(fragment_path, target_path)
def draw_sample(sample_path, out_files, num_samples):
with_protein = (len(out_files) == num_samples + 3)
in_file = out_files[1]
in_sdf = in_file if isinstance(in_file, str) else in_file.name
input_fragments_content = read_molecule_content(in_sdf)
fragments_fmt = in_sdf.split('.')[-1]
offset = 2
input_target_content = None
target_fmt = None
if with_protein:
offset += 1
in_pdb = out_files[2] if isinstance(out_files[2], str) else out_files[2].name
input_target_content = read_molecule_content(in_pdb)
target_fmt = in_pdb.split('.')[-1]
out_sdf = sample_path if isinstance(sample_path, str) else sample_path.name
generated_molecule_content = read_molecule_content(out_sdf)
molecule_fmt = out_sdf.split('.')[-1]
if with_protein:
html = output.SAMPLES_WITH_TARGET_RENDERING_TEMPLATE.format(
fragments=input_fragments_content,
fragments_fmt=fragments_fmt,
molecule=generated_molecule_content,
molecule_fmt=molecule_fmt,
target=input_target_content,
target_fmt=target_fmt,
)
else:
html = output.SAMPLES_RENDERING_TEMPLATE.format(
fragments=input_fragments_content,
fragments_fmt=fragments_fmt,
molecule=generated_molecule_content,
molecule_fmt=molecule_fmt,
)
return output.IFRAME_TEMPLATE.format(html=html)
def compress(output_fnames, name):
archive_path = f'results/all_files_{name}.zip'
with ZipFile(archive_path, 'w') as archive:
for fname in output_fnames:
archive.write(fname)
return archive_path
def generate(in_fragments, in_protein, n_steps, n_atoms, num_samples, selected_atoms):
if in_fragments is None:
return [None, None, None, None]
if in_protein is None:
return generate_without_pocket(in_fragments, n_steps, n_atoms, num_samples, selected_atoms)
else:
return generate_with_pocket(in_fragments, in_protein, n_steps, n_atoms, num_samples, selected_atoms)
def generate_without_pocket(input_file, n_steps, n_atoms, num_samples, selected_atoms):
# Parsing selected atoms (javascript output)
selected_atoms = selected_atoms.strip()
if selected_atoms == '':
selected_atoms = []
else:
selected_atoms = list(map(int, selected_atoms.split(',')))
# Selecting model
if len(selected_atoms) == 0:
selected_model_name = 'geom_difflinker'
else:
selected_model_name = 'geom_difflinker_given_anchors'
print(f'Start generating with model {selected_model_name}, selected_atoms:', selected_atoms)
ddpm = diffusion_models[selected_model_name]
path = input_file.name
extension = path.split('.')[-1]
if extension not in ['sdf', 'pdb', 'mol', 'mol2']:
msg = output.INVALID_FORMAT_MSG.format(extension=extension)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
try:
molecule = read_molecule(path)
try:
molecule = Chem.RemoveAllHs(molecule)
except:
pass
name = '.'.join(path.split('/')[-1].split('.')[:-1])
inp_sdf = f'results/input_{name}.sdf'
except Exception as e:
e = str(e).replace('\'', '')
error = f'Could not read the molecule: {e}'
msg = output.ERROR_FORMAT_MSG.format(message=error)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
if molecule.GetNumAtoms() > 100:
error = f'Too large molecule: upper limit is 100 heavy atoms'
msg = output.ERROR_FORMAT_MSG.format(message=error)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
with Chem.SDWriter(inp_sdf) as w:
w.SetKekulize(False)
w.write(molecule)
positions, one_hot, charges = parse_molecule(molecule, is_geom=True)
anchors = np.zeros_like(charges)
anchors[selected_atoms] = 1
fragment_mask = np.ones_like(charges)
linker_mask = np.zeros_like(charges)
print('Read and parsed molecule')
dataset = [{
'uuid': '0',
'name': '0',
'positions': torch.tensor(positions, dtype=const.TORCH_FLOAT, device=device),
'one_hot': torch.tensor(one_hot, dtype=const.TORCH_FLOAT, device=device),
'charges': torch.tensor(charges, dtype=const.TORCH_FLOAT, device=device),
'anchors': torch.tensor(anchors, dtype=const.TORCH_FLOAT, device=device),
'fragment_mask': torch.tensor(fragment_mask, dtype=const.TORCH_FLOAT, device=device),
'linker_mask': torch.tensor(linker_mask, dtype=const.TORCH_FLOAT, device=device),
'num_atoms': len(positions),
}] * num_samples
dataloader = get_dataloader(dataset, batch_size=num_samples, collate_fn=collate_with_fragment_edges)
print('Created dataloader')
ddpm.edm.T = n_steps
if n_atoms == 0:
def sample_fn(_data):
out, _ = size_nn.forward(_data, return_loss=False)
probabilities = torch.softmax(out, dim=1)
distribution = torch.distributions.Categorical(probs=probabilities)
samples = distribution.sample()
sizes = []
for label in samples.detach().cpu().numpy():
sizes.append(size_nn.linker_id2size[label])
sizes = torch.tensor(sizes, device=samples.device, dtype=torch.long)
return sizes
else:
def sample_fn(_data):
return torch.ones(_data['positions'].shape[0], device=device, dtype=torch.long) * n_atoms
for data in dataloader:
try:
generate_linkers(ddpm=ddpm, data=data, sample_fn=sample_fn, name=name, with_pocket=False)
except Exception as e:
e = str(e).replace('\'', '')
error = f'Caught exception while generating linkers: {e}'
msg = output.ERROR_FORMAT_MSG.format(message=error)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
out_files = try_to_convert_to_sdf(name, num_samples)
out_files = [inp_sdf] + out_files
out_files = [compress(out_files, name=name)] + out_files
choice = out_files[2]
return [
draw_sample(choice, out_files, num_samples),
out_files,
gr.Dropdown.update(
choices=out_files[2:],
value=choice,
visible=True,
),
None
]
def generate_with_pocket(in_fragments, in_protein, n_steps, n_atoms, num_samples, selected_atoms):
# Parsing selected atoms (javascript output)
selected_atoms = selected_atoms.strip()
if selected_atoms == '':
selected_atoms = []
else:
selected_atoms = list(map(int, selected_atoms.split(',')))
# Selecting model
if len(selected_atoms) == 0:
selected_model_name = 'pockets_difflinker'
else:
selected_model_name = 'pockets_difflinker_given_anchors'
print(f'Start generating with model {selected_model_name}, selected_atoms:', selected_atoms)
ddpm = diffusion_models[selected_model_name]
fragments_path = in_fragments.name
fragments_extension = fragments_path.split('.')[-1]
if fragments_extension not in ['sdf', 'pdb', 'mol', 'mol2']:
msg = output.INVALID_FORMAT_MSG.format(extension=fragments_extension)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
protein_path = in_protein.name
protein_extension = protein_path.split('.')[-1]
if protein_extension not in ['pdb']:
msg = output.INVALID_FORMAT_MSG.format(extension=protein_extension)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
try:
fragments_mol = read_molecule(fragments_path)
name = '.'.join(fragments_path.split('/')[-1].split('.')[:-1])
except Exception as e:
e = str(e).replace('\'', '')
error = f'Could not read the molecule: {e}'
msg = output.ERROR_FORMAT_MSG.format(message=error)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
if fragments_mol.GetNumAtoms() > 100:
error = f'Too large molecule: upper limit is 100 heavy atoms'
msg = output.ERROR_FORMAT_MSG.format(message=error)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
inp_sdf = f'results/input_{name}.sdf'
with Chem.SDWriter(inp_sdf) as w:
w.SetKekulize(False)
w.write(fragments_mol)
inp_pdb = f'results/target_{name}.pdb'
shutil.copy(protein_path, inp_pdb)
frag_pos, frag_one_hot, frag_charges = parse_molecule(fragments_mol, is_geom=True)
pocket_pos, pocket_one_hot, pocket_charges = get_pocket(fragments_mol, protein_path)
print(f'Detected pocket with {len(pocket_pos)} atoms')
positions = np.concatenate([frag_pos, pocket_pos], axis=0)
one_hot = np.concatenate([frag_one_hot, pocket_one_hot], axis=0)
charges = np.concatenate([frag_charges, pocket_charges], axis=0)
anchors = np.zeros_like(charges)
anchors[selected_atoms] = 1
fragment_only_mask = np.concatenate([
np.ones_like(frag_charges),
np.zeros_like(pocket_charges),
])
pocket_mask = np.concatenate([
np.zeros_like(frag_charges),
np.ones_like(pocket_charges),
])
linker_mask = np.concatenate([
np.zeros_like(frag_charges),
np.zeros_like(pocket_charges),
])
fragment_mask = np.concatenate([
np.ones_like(frag_charges),
np.ones_like(pocket_charges),
])
print('Read and parsed molecule')
dataset = [{
'uuid': '0',
'name': '0',
'positions': torch.tensor(positions, dtype=const.TORCH_FLOAT, device=device),
'one_hot': torch.tensor(one_hot, dtype=const.TORCH_FLOAT, device=device),
'charges': torch.tensor(charges, dtype=const.TORCH_FLOAT, device=device),
'anchors': torch.tensor(anchors, dtype=const.TORCH_FLOAT, device=device),
'fragment_only_mask': torch.tensor(fragment_only_mask, dtype=const.TORCH_FLOAT, device=device),
'pocket_mask': torch.tensor(pocket_mask, dtype=const.TORCH_FLOAT, device=device),
'fragment_mask': torch.tensor(fragment_mask, dtype=const.TORCH_FLOAT, device=device),
'linker_mask': torch.tensor(linker_mask, dtype=const.TORCH_FLOAT, device=device),
'num_atoms': len(positions),
}] * num_samples
dataset = MOADDataset(data=dataset)
ddpm.val_dataset = dataset
batch_size = min(num_samples, MAX_BATCH_SIZE)
dataloader = get_dataloader(dataset, batch_size=batch_size, collate_fn=collate_with_fragment_without_pocket_edges)
print('Created dataloader')
ddpm.edm.T = n_steps
if n_atoms == 0:
def sample_fn(_data):
out, _ = size_nn.forward(_data, return_loss=False, with_pocket=True)
probabilities = torch.softmax(out, dim=1)
distribution = torch.distributions.Categorical(probs=probabilities)
samples = distribution.sample()
sizes = []
for label in samples.detach().cpu().numpy():
sizes.append(size_nn.linker_id2size[label])
sizes = torch.tensor(sizes, device=samples.device, dtype=torch.long)
return sizes
else:
def sample_fn(_data):
return torch.ones(_data['positions'].shape[0], device=device, dtype=torch.long) * n_atoms
for batch_i, data in enumerate(dataloader):
try:
offset_idx = batch_i * batch_size
generate_linkers(
ddpm=ddpm, data=data,
sample_fn=sample_fn, name=name, with_pocket=True,
offset_idx=offset_idx,
)
except Exception as e:
e = str(e).replace('\'', '')
error = f'Caught exception while generating linkers: {e}'
msg = output.ERROR_FORMAT_MSG.format(message=error)
return [output.IFRAME_TEMPLATE.format(html=msg), None, None, None]
out_files = try_to_convert_to_sdf(name, num_samples)
out_files = [inp_sdf, inp_pdb] + out_files
out_files = [compress(out_files, name=name)] + out_files
choice = out_files[3]
return [
draw_sample(choice, out_files, num_samples),
out_files,
gr.Dropdown.update(
choices=out_files[3:],
value=choice,
visible=True,
),
None
]
demo = gr.Blocks()
with demo:
gr.Markdown('# DiffLinker: Equivariant 3D-Conditional Diffusion Model for Molecular Linker Design')
gr.Markdown(
'Given a set of disconnected fragments in 3D, '
'DiffLinker places missing atoms in between and designs a molecule incorporating all the initial fragments. '
'Our method can link an arbitrary number of fragments, requires no information on the attachment atoms '
'and linker size, and can be conditioned on the protein pockets.'
)
gr.Markdown(
'[**[Paper]**](https://arxiv.org/abs/2210.05274) '
'[**[Code]**](https://github.com/igashov/DiffLinker)'
)
with gr.Box():
with gr.Row():
with gr.Column():
gr.Markdown('## Input')
gr.Markdown('Upload the file with 3D-coordinates of the input fragments in .pdb, .mol2 or .sdf format:')
input_fragments_file = gr.File(file_count='single', label='Input Fragments')
gr.Markdown('Upload the file of the target protein in .pdb format (optionally):')
input_protein_file = gr.File(file_count='single', label='Target Protein (Optional)')
n_steps = gr.Slider(
minimum=MIN_N_STEPS, maximum=MAX_N_STEPS,
label="Number of Denoising Steps", step=10
)
n_atoms = gr.Slider(
minimum=0, maximum=20,
label="Linker Size: DiffLinker will predict it if set to 0",
step=1
)
n_samples = gr.Slider(minimum=5, maximum=50, label="Number of Samples", step=5)
examples = gr.Dataset(
components=[gr.File(visible=False), gr.File(visible=False)],
samples=[
['examples/example_1.sdf', ''],
['examples/example_2.sdf', ''],
['examples/3hz1_fragments.sdf', 'examples/3hz1_protein.pdb'],
['examples/5ou2_fragments.sdf', 'examples/5ou2_protein.pdb'],
],
type='values',
headers=['Input Fragments', 'Target Protein'],
)
button = gr.Button('Generate Linker!')
gr.Markdown('')
gr.Markdown('## Output Files')
gr.Markdown('Download files with the generated molecules here:')
output_files = gr.File(file_count='multiple', label='Output Files', interactive=False)
hidden = gr.Textbox(visible=False)
with gr.Column():
gr.Markdown('## Visualization')
gr.Markdown('**Hint:** click on atoms to select anchor points (optionally)')
samples = gr.Dropdown(
choices=[],
value=None,
type='value',
multiselect=False,
visible=False,
interactive=True,
label='Samples'
)
visualization = gr.HTML()
input_fragments_file.change(
fn=show_input,
inputs=[input_fragments_file, input_protein_file],
outputs=[visualization, samples, hidden],
)
input_protein_file.change(
fn=show_input,
inputs=[input_fragments_file, input_protein_file],
outputs=[visualization, samples, hidden],
)
input_fragments_file.clear(
fn=clear_fragments_input,
inputs=[input_protein_file],
outputs=[input_fragments_file, visualization, samples, hidden],
)
input_protein_file.clear(
fn=clear_protein_input,
inputs=[input_fragments_file],
outputs=[input_protein_file, visualization, samples, hidden],
)
examples.click(
fn=click_on_example,
inputs=[examples],
outputs=[input_fragments_file, input_protein_file, visualization, samples, hidden]
)
button.click(
fn=generate,
inputs=[input_fragments_file, input_protein_file, n_steps, n_atoms, n_samples, hidden],
outputs=[visualization, output_files, samples, hidden],
_js=output.RETURN_SELECTION_JS,
)
samples.select(
fn=draw_sample,
inputs=[samples, output_files, n_samples],
outputs=[visualization],
)
demo.load(_js=output.STARTUP_JS)
demo.launch(server_name=args.ip)