pipeline.py

# Copyright 2021 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from typing import Mapping, List
from kfp.v2 import dsl

from src import config

from src.components.jackhmmer import jackhmmer as JackhmmerOp
from src.components.hhblits import hhblits as HhblitsOp
from src.components.hhsearch import hhsearch as HhsearchOp
from src.components.aggregate_features import aggregate_features as AggregateOp
from src.components.model_predict import predict as PredictOp
from src.components.relax_protein import relax as RelaxOp


@dsl.pipeline(
    name=config.PIPELINE_NAME, 
    description=config.PIPELINE_DESCRIPTION
)
def alphafold_pipeline(
    sequence_path: str,
    project: str='alphafold-dev',
    region: str='us-central1',
    max_template_date: str='2020-05-14',
    models: List[Mapping]=[{'model_name': 'model_1', 'random_seed': 1}],
    use_gpu_for_relaxation: bool=True,
    num_ensemble: int=1,
    reference_datasets_uri: str=config.REFERENCE_DATASETS_URI, 
    model_params_uri: str=config.MODEL_PARAMS_GCS_LOCATION
):
    """Runs AlphaFold inference."""

    input_sequence = dsl.importer(
        artifact_uri=sequence_path,
        artifact_class=dsl.Dataset,
        reimport=True)
    input_sequence.set_display_name('Input sequence')

    model_parameters = dsl.importer(
        artifact_uri=model_params_uri,
        artifact_class=dsl.Artifact,
        reimport=True)
    model_parameters.set_display_name('Model parameters')

    reference_databases = dsl.importer(
        artifact_uri=reference_datasets_uri,
        artifact_class=dsl.Dataset,
        reimport=False,
        metadata={
            config.UNIREF90: config.UNIREF90_PATH,
            config.MGNIFY: config.MGNIFY_PATH,
            config.BFD: config.BFD_PATH,
            config.UNICLUST30: config.UNICLUST30_PATH,
            config.PDB70: config.PDB70_PATH,
            config.PDB_MMCIF: config.PDB_MMCIF_PATH,
            config.PDB_OBSOLETE: config.PDB_OBSOLETE_PATH,
            config.PDB_SEQRES: config.PDB_SEQRES_PATH,
            config.UNIPROT: config.UNIPROT_PATH,
            }    
    )
    reference_databases.set_display_name('Reference databases')

    search_uniref = JackhmmerOp(
        project=project,
        region=region,
        database=config.UNIREF90,
        reference_databases=reference_databases.output,
        sequence=input_sequence.output,
        components_image=config.ALPHAFOLD_COMPONENTS_IMAGE
    )
    search_uniref.set_display_name('Search Uniref')#.set_caching_options(enable_caching=True)

    search_mgnify = JackhmmerOp(
        project=project,
        region=region,
        database=config.MGNIFY,
        reference_databases=reference_databases.output,
        sequence=input_sequence.output,
        components_image=config.ALPHAFOLD_COMPONENTS_IMAGE
    )
    search_mgnify.set_display_name('Search Mgnify')#.set_caching_options(enable_caching=True)

    search_uniclust = HhblitsOp(
        project=project,
        region=region,
        msa_dbs=[config.UNICLUST30],
        reference_databases=reference_databases.output,
        sequence=input_sequence.output,
        components_image=config.ALPHAFOLD_COMPONENTS_IMAGE
    )
    search_uniclust.set_display_name('Search Uniclust')#.set_caching_options(enable_caching=True)

    search_bfd = HhblitsOp(
        project=project,
        region=region,
        msa_dbs=[config.BFD],
        reference_databases=reference_databases.output,
        sequence=input_sequence.output,
        components_image=config.ALPHAFOLD_COMPONENTS_IMAGE
    )
    search_bfd.set_display_name('Search BFD')#.set_caching_options(enable_caching=True)

    search_pdb = HhsearchOp(
        project=project,
        region=region,
        template_dbs=[config.PDB70],
        mmcif_db=config.PDB_MMCIF,
        obsolete_db=config.PDB_OBSOLETE,
        max_template_date=max_template_date,
        reference_databases=reference_databases.output,
        sequence=input_sequence.output,
        msa=search_uniref.outputs['msa'],
        components_image=config.ALPHAFOLD_COMPONENTS_IMAGE
    )
    search_pdb.set_display_name('Search Pdb')#.set_caching_options(enable_caching=True)

    aggregate_features = AggregateOp(
        sequence=input_sequence.output,
        msa1=search_uniref.outputs['msa'],
        msa2=search_mgnify.outputs['msa'],
        msa3=search_bfd.outputs['msa'],
        msa4=search_uniclust.outputs['msa'],
        template_features=search_pdb.outputs['template_features'],
    )
    aggregate_features.set_display_name('Aggregate features')#.set_caching_options(enable_caching=True)

    # Think what to do with random seed when switch to Parallel loop
    with dsl.ParallelFor(models) as model:
        model_predict = PredictOp(
            model_features=aggregate_features.outputs['features'],
            model_params=model_parameters.output,
            model_name=model.model_name,
            num_ensemble=num_ensemble,
            random_seed=model.random_seed
        )
        model_predict.set_display_name('Predict')#.set_caching_options(enable_caching=True)
        model_predict.set_cpu_limit(config.CPU_LIMIT)
        model_predict.set_memory_limit(config.MEMORY_LIMIT)
        model_predict.set_gpu_limit(config.GPU_LIMIT)
        model_predict.add_node_selector_constraint(config.GKE_ACCELERATOR_KEY, config.GPU_TYPE)
        model_predict.set_env_variable("TF_FORCE_UNIFIED_MEMORY", config.TF_FORCE_UNIFIED_MEMORY)
        model_predict.set_env_variable("XLA_PYTHON_CLIENT_MEM_FRACTION", config.XLA_PYTHON_CLIENT_MEM_FRACTION)

        relax_protein = RelaxOp(
            unrelaxed_protein=model_predict.outputs['unrelaxed_protein'],
            use_gpu=use_gpu_for_relaxation,
        )
        relax_protein.set_display_name('Relax protein')#.set_caching_options(enable_caching=True)
        relax_protein.set_cpu_limit(config.RELAX_CPU_LIMIT)
        relax_protein.set_memory_limit(config.RELAX_MEMORY_LIMIT)
        relax_protein.set_gpu_limit(config.RELAX_GPU_LIMIT)
        relax_protein.add_node_selector_constraint(config.GKE_ACCELERATOR_KEY, config.RELAX_GPU_TYPE)
        relax_protein.set_env_variable("TF_FORCE_UNIFIED_MEMORY", config.TF_FORCE_UNIFIED_MEMORY)
        relax_protein.set_env_variable("XLA_PYTHON_CLIENT_MEM_FRACTION", config.XLA_PYTHON_CLIENT_MEM_FRACTION)