Overview - Data

Get an overview of using data at Cradle.

Import your experimental data to Cradle to use in machine learning tasks. You can start with existing data or begin generating new data through Cradle's design tasks.

As you work with Cradle you generate sequences with various tasks, test them in the lab and import the assay data the platform. By importing your data Cradle's models will be able to learn and improve to better understand your protein.

Using data at Cradle

1. 1

   Create a table as a data schema

   Create a table that defines the structure and format of your experimental data before importing. Tables serve as schemas that ensure consistent data organization across your workspace.

   Before you import data, define a table that describes the structure of the data you plan to upload. This is either defined by a workspace administrator (to keep data imports consistent across a company), or you can define it yourself. You can create different tables for different use cases.

   Learn more in tables.

2. 2

   Evaluate data

   Your data's quality and quantity directly impact model performance. High-quality datasets with appropriate controls enable better learning outcomes for your protein engineering projects.

   To evaluate your data run the Analyze data task. This will produce a report on the quality and quantity of your data for machine learning. Learn more in /task/create. Read more about best practices for your data in data guidelines.

3. 3

   Import files

   When importing data to Cradle you append to a table. Cradle currently accepts .csv, .tsv and .xlsx filetypes. If you upload through the API you can additionally upload .parquet formats. You can add multiple imports to a table and track its version history over time.

   Learn more in importing data.

4. 4

   Create views

   Views are saved queries that filter and organize your table data for specific use cases. Use views to prepare data inputs for different tasks without modifying your underlying tables.

   Learn more in views.