All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Updated S3 location for reference data
- Remove jax from notebook requirements
- Update AlphaFold to version v2.3.2
- Removed VPC configuration for Lambda-based custom resources to better support bring-you-own-VPC
- Updated scipy version in JackHMMER container in response to Dependabot alert.
- Optional p4d compute environment and job queue
- Deployment script for updating existing stacks
- Updating the cfn stack will now restart the CodeBuild jobs for all module containers
- Fixed bug in miniconda version for OpenFold, ESMFold, and NextFlow container images.
- Updated ESMFold container to match Amazon HealthOmics implementation and improve run outputs.
- Due to a lack of ARM in the HMMER package, reverting all non-accelerated jobs back to x86 architecture.
- Support for running AlphaFold jobs on non-accelerated instance types by setting
gpu=0
and submitting job to the CPUOnDemandJobQueue. - Support for passing a dictionary of environment variables to AlphaFold jobs.
- Support for c/m/r6i instance types in the CPU on-demand compute environment.
- Support for c/m/r7g instance types on the Graviton on-demand and spot compute environments.
- Added model version information to README.
- Updated AlphaFold CUDA version to 11.6
- Fixed bug causing hhsearch errors on g5 instance types
- Fixed bug preventing load of nested stacks.
- Added additional error handling to batchfold package.
- Added paginators to boto3 calls in batchfold environment class to address issue with deploying multiple stacks.
- Added support for NextFlow
- Added support for RFDiffusion
- Removed deprecated notebook for RFDesign
- Update ProteinMPNN container and job definition to use accelerated instance types.
- Add new option to ProteinMPNN inference script to remove the input sequence from the output
- Fixed bug with ESMFold metric logging for multiple sequence inputs.
- Added support for DiffDock
- Added ECR URIs as outputs to the CloudFormation template
- Fixed typo in the pip install cell of the AlphaFold-Multimer notebook
- Updated version of py3Dmol to address structure visualization issue in JupyterLab
- Pin base image for ProteinMPNN and download containers.
- Check for AWSServiceRoleForEC2Spot and AWSServiceRoleForEC2SpotFleet service-linked roles before creating duplicates
- Updated AlphaFold version to 2.3.1
- Added support for OmegaFold release 2 models
- Moved reference data to S3 for faster stack provisioning
- Pinned commit of OmegaFold repository
- Fixed issue with ESMFold Dockerfile
- Added support for ProteinMPNN jobs.
- Incorporated ProteinMPNN into protein design notebook.
- Added list of supported algorithms to README index.
- Fixed ESMFold issue with malformed fasta headers
- Pinned ESMFold dependencies
- Added support for ESMFold jobs.
- Added support for RFDesign hallucination and inpainting jobs.
- Added default support for multi-AZ deployments.
- Added an S3 data repository to FSx for Lustre deployment to back up reference data.
- Added exception handling for invalid JackHMMER
db_preset
values.
- Improved clarity of
prep_databases.py
script. - Removed Batch compute environments for amd instance types.
- Moved S3 resource to root CloudFormation stack and updated retention policy
- Updated cost estimates.
- Support for the OmegaFold LLM-based protein folding algorithm.
- Resolved issue with plotting protein structures on SageMaker Notebook instances.
- Update AlphaFold version to v2.2.4.
- Added nbhelper functions to BatchFold package
- Improved README
- Refactor code to make algorithms more modular. This will make it easier to add new ones!
- Initial Release