Skip to main content

First round with Cradle

Learn how to get started with Cradle and go from creating your first project to generating candidates to test in the lab.

  1. 1

    Set up your account

    Connect to a Customer Success Manager at Cradle or your workspace administrator to get an invitation to join Cradle. Once you receive an invitation to Cradle by email, you can log in using Single Sign-On (SSO) with your organization's email address.

  2. 2

    Create your first project

    To get started, create a project for the protein you are trying to optimize. Within a project you import data, run tasks for your first round to generate sequences and analyze results through reports generate sequences and analyze results through reports.

    Learn more in Projects.

  3. 3

    Import data

    You can get started with Cradle with or without any experimental data. If you have existing experimental data from the lab for your project, make sure the quality and quantity is sufficient for machine learning models to learn from. Learn more in data guidelines.

    To start importing data you create a table, a well-defined column schema for your data imports. After adding a table, import the files you want to add to your project.

    Learn more in Data.

  4. 4

    Set up a round

    Within your project, start your first task and assign it to a new round. One round typically includes experiments related to a specific sub-goal or time frame. Within a round, you can run many tasks to interact with Cradle's ML.

    Projects can include several rounds, each serving as a grouping for when you design sequences and test them in the lab.

    Learn more in Rounds.

  5. 5

    Run a task

    You use Tasks to interact with Cradle's machine learning. You have control over which Tasks to run and how you configure them. Learn more in Tasks.

    When starting with Cradle you typically begin by uploading historical data and running an “Analyse data” task to get information about how well your data will work to train models. If the data analysis indicates that your data looks good, you then run a “Train” task to start training a model on your protein of interest. When you are satisfied with models, you can run the “Engineer” task to engineer new sequence candidates to take to the lab.

    Learn more about what Tasks you can run in the API documentation for /task/create.

  6. 6

    Viewing reports

    Most long running tasks output analysis to help you understand the results of a task. For example the Analyze Data task shows you how well your data is suited to training an ML model. The Train task outputs analysis on the models trained and the correlation between assays. The Generate sequences task outputs analysis on the generated sequences. This analysis is combined into a Report and can be viewed by anyone in your workspace.

  7. 7

    Testing in the lab

    Test the sequences in your own lab or through an outsourced service. After testing, upload the results to the Cradle platform. This data will further improve your Cradle machine learning models and enable you to generate better sequences in the next round.

Congrats! You ran your first round with Cradle. As you continue on your project, the models the platform created for you will continue to understand the protein you are trying to optimize. Round over round, the variants you generate with Cradle will get better and better.

Last updated on