Overview#

Metasmith provides a type system to describe bioinformatics data products based on how it can be generated or consumed by computational tools. This enables a solver to generate Nextflow workflows from target types, given data, and available tools. Generated workflows can then be executed directly or on remote machines using containerization.

Install Metasmith

Usage in 5 steps#

1 - Deploy an agent#

Metasmith itself is not an agent. Rather, Metasmith spawns and deploys agents to machines on which you intend to execute workflows. Agent deployment is automated; you do not need to install Metasmith again if executing workflows on remote machines.

smith = Agent(...)
smith.Deploy()

2 - Register inputs#

Metasmith uses a type system to model how data can be consumed by protocols. Inputs must be registered by assigning a type to them.

inputs = DataInstanceLibrary(...)
inputs.AddItem(
    <file path>,
    <data type>,
)

3 - Generate workflow#

Protocols describe transformations between data types. By providing the inputs and a list of available transformations, we can ask the agent to generate a workflow to produce target data types by chaining multiple protocols together.

task = smith.GenerateWorkflow(
    <inputs>,
    <transforms>,
    <targets>,
)

Tip

Browse the standard library of transforms.

4 - Execute workflow#

Staging a workflow translates it into Nextflow’s syntax and prepares default configurations for various platforms. When ready, execution is delegated to Nextflow.

smith.StageWorkflow(task)
smith.RunWorkflow(task)
smith.CheckWorkflow(task)

5 - Receive outputs#

Produced outputs are presented as a structured data product, but exists as a simple folder to maintain accessibility by both humans and machines.

outputs = DataInstanceLibrary.LoadFrom(
    smith.GetResultSource(task),
    ...
)