Predicting sequences

This tutorial shows you how to get the results of your predict jobs when using Property Regression models. Use this to predict scores for new sequences using models trained on your old sequences.

What you need before getting started

You need a trained model. See Training models for more information.

Predicting your sequences

Initiate your predict job with new sequences of interest:

new_sequences = ["PPPGDISSSNDTVGVAVVNYKMPRLHTAAEVLDNARKIAEMIVGMKQGLPGMDLVVFPEYSLQGIMYDPAEMMETAVAIPGEETEIFSRACRKANVWGVFSLTGERHEEHPRKAPYNTLVLIDNNGEIVQKYRKIIPWCPIEGWYPGGQTYVSEGPKGMKISLIICDDGNYPEIWRDCAMKGAELIVRCQGYMYPAKDQQVMMAKAMAWANNCYVAVANAAGFDGVYSYFGHSAIIGFDGRTLGECGEEEMGIQYAQLSLSQIRDARANDQSQNHLFKILHRGYSGLQASGDGDRGLAECPFEFYRTWVTDAEKARENVERLTRSTTGVAQCPVGRLPYEGLEKEA",
                 "GGGGDISSSNDTVGVAVVNYKMPRLHTAAEVLDNARKIAEMIVGMKQGLPGMDLVVFPEYSLQGIMYDPAEMMETAVAIPGEETEIFSRACRKANVWGVFSLTGERHEEHPRKAPYNTLVLIDNNGEIVQKYRKIIPWCPIEGWYPGGQTYVSEGPKGMKISLIICDDGNYPEIWRDCAMKGAELIVRCQGYMYPAKDQQVMMAKAMAWANNCYVAVANAAGFDGVYSYFGHSAIIGFDGRTLGECGEEEMGIQYAQLSLSQIRDARANDQSQNHLFKILHRGYSGLQASGDGDRGLAECPFEFYRTWVTDAEKARENVERLTRSTTGVAQCPVGRLPYEGLEKEA"]

pjob = predictor.predict(sequences=new_sequences)
pjob

PredictJob(job_id='6031f2b7-cf98-4b92-b8d8-ffb72f619d7c', job_type=<JobType.predictor_predict: '/predictor/predict'>, status=<JobStatus.PENDING: 'PENDING'>, created_date=datetime.datetime(2024, 12, 3, 5, 40, 46, 933939, tzinfo=TzInfo(UTC)), start_date=None, end_date=None, prerequisite_job_id=None, progress_message=None, progress_counter=0, sequence_length=None)

View your results when the job is complete:

results = pjob.wait(verbose=True)

Waiting: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 100/100 [00:01<00:00, 63.41it/s, status=SUCCESS]

from collections import namedtuple
Score = namedtuple("Score", ["sequence", "mean", "var"])
[Score(s, mu.item(), var.item()) for s, mu, var in zip(pjob.sequences, results[0].T[0], results[1].T[0])]

[Score(sequence=b'PPPGDISSSNDTVGVAVVNYKMPRLHTAAEVLDNARKIAEMIVGMKQGLPGMDLVVFPEYSLQGIMYDPAEMMETAVAIPGEETEIFSRACRKANVWGVFSLTGERHEEHPRKAPYNTLVLIDNNGEIVQKYRKIIPWCPIEGWYPGGQTYVSEGPKGMKISLIICDDGNYPEIWRDCAMKGAELIVRCQGYMYPAKDQQVMMAKAMAWANNCYVAVANAAGFDGVYSYFGHSAIIGFDGRTLGECGEEEMGIQYAQLSLSQIRDARANDQSQNHLFKILHRGYSGLQASGDGDRGLAECPFEFYRTWVTDAEKARENVERLTRSTTGVAQCPVGRLPYEGLEKEA', mean=-0.415189117193222, var=0.1528152972459793),
 Score(sequence=b'GGGGDISSSNDTVGVAVVNYKMPRLHTAAEVLDNARKIAEMIVGMKQGLPGMDLVVFPEYSLQGIMYDPAEMMETAVAIPGEETEIFSRACRKANVWGVFSLTGERHEEHPRKAPYNTLVLIDNNGEIVQKYRKIIPWCPIEGWYPGGQTYVSEGPKGMKISLIICDDGNYPEIWRDCAMKGAELIVRCQGYMYPAKDQQVMMAKAMAWANNCYVAVANAAGFDGVYSYFGHSAIIGFDGRTLGECGEEEMGIQYAQLSLSQIRDARANDQSQNHLFKILHRGYSGLQASGDGDRGLAECPFEFYRTWVTDAEKARENVERLTRSTTGVAQCPVGRLPYEGLEKEA', mean=-0.4151889681816101, var=0.1528152972459793)]

Next steps:

Visit the Predictor API page to learn more.