openprotein-python

The OpenProtein.AI Python Interface provides a user-friendly library to interact with the OpenProtein.AI REST API, enabling various tasks related to protein analysis and modeling.

Table of Contents

	Workflow	Description
0	Quick start	Quick start guide
1	Installation	Install guide for pip and conda.
2	Session management	An overview of the OpenProtein Python Client & the asynchronous jobs system.
3	Asssay-based Sequence Learning	Covers core tasks such as data upload, model training & prediction, and sequence design.
4	De Novo prediction & generative models (PoET)	Covers PoET, a protein LLM for de novo scoring, as well as sequence generation.
5	Protein Language Models & Embeddings	Covers methods for creating sequence embeddings with proprietary & open-source models.

Quick-start

Get started with our quickstart README! You can peruse the official documentation for more details!

Installation

To install the python interface using pip, run the following command:

pip install openprotein-python

or with conda:

conda install -c openprotein openprotein-python

Requirements

• Python 3.8 or higher.
• pydantic version 1.0 or newer.
• requests version 2.0 or newer.
• tqdm version 4.0 or newer.
• pandas version 1.0 or newer.

Getting started

Read on below for the quick-start guide, or see the docs for more information!

To begin, create a session using your login credentials.

import openprotein

# replace USERNAME and PASSWORD with your actual login credentials
session = openprotein.connect(USERNAME, PASSWORD)

Job Status

The interface offers AsyncJobFuture objects for asynchronous calls, allowing tracking of job status and result retrieval when ready. Given a future, you can check its status and retrieve results.

Checking Job Status

Check the status of an AsyncJobFuture using the following methods:

future.refresh()  # call the backend to update the job status
future.done()     # returns True if the job is done, meaning the status could be SUCCESS, FAILED, or CANCELLED

Retrieving Job Results

Once the job has finished, retrieve the results using the following methods:

result = future.wait()     # wait until done and then fetch results

#verbosity is controlled with verbose arg
result = future.get(verbose=True)  # get the result from a finished job

Jobs Interface

Listing Jobs

To view all jobs associated with each session, the following method is available, providing an option to filter results by date, job type, or status.

session.jobs.list() 

Retrieving Specific Job

For detailed information about a particular job, use the following command with the corresponding job ID:

session.jobs.get(JOB_ID)  # Replace JOB_ID with the ID of the specific job to be retrieved

Resuming Jobs

Jobs from prior workflows can be resumed using the load_job method provided by each API.

session.load_job(JOB_ID)  # Replace JOB_ID with the ID of the training job to resume

PoET interface

The PoET Interface allows scoring, generating, and retrieving sequences using the PoET model.

Scoring Sequences

To score sequences, use the score function. Provide a prompt and a list of queries. The results will be a list of (sequence, score) pydantic objects.

prompt_seqs = b'MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN'

prompt = session.poet.upload_prompt(prompt_seqs)

queries = [
    b'MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN',
    b'MALWMRLLPLLVLLALWGPDPASAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN',
    b'MALWTRLRPLLALLALWPPPPARAFVNQHLCGSHLVEALYLVCGERGFFYTPKARREVEGPQVGALELAGGPGAGGLEGPPQKRGIVEQCCASVCSLYQLENYCN',
    b'MALWIRSLPLLALLVFSGPGTSYAAANQHLCGSHLVEALYLVCGERGFFYSPKARRDVEQPLVSSPLRGEAGVLPFQQEEYEKVKRGIVEQCCHNTCSLYQLENYCN',
    b'MALWMRLLPLLALLALWAPAPTRAFVNQHLCGSHLVEALYLVCGERGFFYTPKARREVEDLQVRDVELAGAPGEGGLQPLALEGALQKRGIVEQCCTSICSLYQLENYCN',
]

future = session.poet.score(prompt, queries)
result = future.wait()
# result is a list of (sequence, score) pydantic objects

Scoring Single Site Variants

For scoring single site variants, use the single_site function, providing the original sequence and setting prompt_is_seed to True if the prompt is a seed sequence.

sequence = "MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN"
future = session.poet.single_site(prompt, sequence, prompt_is_seed=True) 
result = future.wait()
# result is a dictionary of {variant: score}

Generating Sequences

To generate sequences from the PoET model, use the generate function with relevant parameters. The result will be a list of generated samples.

future = session.poet.generate(
    prompt,
    max_seqs_from_msa=1024,
    num_samples=100,
    temperature=1.0,
    topk=15
)
samples = future.wait()

Retrieving Input Sequences

You can retrieve the prompt, MSA, or seed sequences for a PoET job using the get_input function or the individual functions for each type.

future.get_input(INPUT_TYPE)
# or, functions for each type
future.get_prompt()
future.get_msa()
future.get_seed()

See more at our Homepage