Alphapulldown

🚀 Automated Workflows with Snakemake
Introduced a streamlined pipeline for modeling tasks, leveraging Snakemake for automation, scalability, and reproducibility. Installation is now simplified with Docker/Singularity containers.

🔧 Unified Configuration Syntax
All AlphaPulldown modes now share a unified configuration syntax, simplifying input setup and enabling automatic sequence retrieval with UniProt IDs.

🛠️ Modular Backend Support
Reorganized the codebase to support multiple modeling backends flexibly. UniFold and AlphaLink2 are now integrated, with the potential to add more in the future.

🔗 Cross-Link-Driven Modeling
AlphaLink2 integration supports modeling with cross-linking mass spectrometry (XL-MS) data, enhancing the accuracy of complex structural models.

💾 Significant Storage Optimization
Achieved more than 90% reduction in storage through input feature and output file compression, promoting sustainability in large-scale modeling.

📦 ModelCIF Format Support
Models can now be stored in ModelCIF format, aligning with FAIR principles for improved model accessibility and simpler model deposition in databases.

🧩 Extended Modeling Capabilities
Increased flexibility with customizable modeling parameters, multimeric template support, and options to control MSA and template impact on predictions.

📊 Enhanced Analysis Pipeline
Enriched the analysis toolkit with additional evaluation metrics like average pLDDT and PAE scores at protein interfaces for better assessment of model confidence.

🔄 Improved Codebase and Documentation
Refactored code, introduced CI/CD pipelines, automated testing, and expanded documentation for a smoother user experience.

🌐 Repository of Precomputed Features
Released a web-based [repository of precomputed input features](https://github.com/KosinskiLab/AlphaPulldown?tab=readme-ov-file#features-database) for multiple model organisms, reducing redundant computations and accelerating workflows.

🚀 Automated Workflows with Snakemake
Introduced a streamlined pipeline for modeling tasks, leveraging **Snakemake** for automation, scalability, and reproducibility. Installation is now simplified with **Docker/Singularity** containers.

🔧 Unified Configuration Syntax
All AlphaPulldown modes now share a **unified configuration syntax**, simplifying input setup and enabling automatic sequence retrieval with **UniProt IDs**.

🛠️ Modular Backend Support
Reorganized the codebase to support multiple modeling backends flexibly. **UniFold** and **AlphaLink2** are now integrated, with the potential to add more in the future.

🔗 Cross-Link-Driven Modeling
**AlphaLink2** integration supports modeling with cross-linking mass spectrometry (**XL-MS**) data, enhancing the accuracy of complex structural models.

💾 Significant Storage Optimization
Achieved more than **90% reduction** in storage through input feature and output file compression, promoting sustainability in large-scale modeling.

📦 ModelCIF Format Support
Models can now be stored in **ModelCIF format**, aligning with FAIR principles for improved model accessibility and simpler model deposition in databases.

🧩 Extended Modeling Capabilities
Increased flexibility with customizable modeling parameters, multimeric template support, and options to control **MSA** and template impact on predictions.

📊 Enhanced Analysis Pipeline
Enriched the analysis toolkit with additional evaluation metrics like average **pLDDT** and **PAE** scores at protein interfaces for better assessment of model confidence.

🔄 Improved Codebase and Documentation
Refactored code, introduced **CI/CD pipelines**, automated testing, and expanded documentation for a smoother user experience.

🌐 Repository of Precomputed Features
Released a web-based [**repository of precomputed input features**](https://github.com/KosinskiLab/AlphaPulldown?tab=readme-ov-file#features-database) for multiple model organisms, reducing redundant computations and accelerating workflows.

---

Release Notes

---

**🚀 Automated Workflows with Snakemake**
Introduced a streamlined pipeline for modeling tasks, leveraging Snakemake for automation, scalability, and reproducibility. Installation is now simplified with Docker/Singularity containers.

**🔧 Unified Configuration Syntax**
All AlphaPulldown modes now share a unified configuration syntax, simplifying input setup and enabling automatic sequence retrieval with UniProt IDs.

**🛠️ Modular Backend Support**
Reorganized codebase to support multiple modeling backends flexibly. UniFold and AlphaLink2 are now integrated, with the potential to add more in the future.

**🔗 Cross-Link-Driven Modeling**
AlphaLink2 integration supports modeling with cross-linking mass spectrometry (XL-MS) data, enhancing the accuracy of complex structural models.

**💾 Significant Storage Optimization**
Achieved more than 90% reduction in storage through input feature and output file compression, promoting sustainability in large-scale modeling.

**📦 ModelCIF Format Support**
Models can now be stored in ModelCIF format, aligning with FAIR principles for improved model accessibility and simpler model deposition in databases.

**🧩 Extended Modeling Capabilities**
Increased flexibility with customizable modeling parameters, multimeric template support, and options to control MSA and template impact on predictions.

**📊 Enhanced Analysis Pipeline**
Enriched the analysis toolkit with additional evaluation metrics like average pLDDT and PAE scores at protein interfaces for better assessment of model confidence..

**🔄 Improved Codebase and Documentation**
Refactored code, introduced CI/CD pipelines, automated testing, and expanded documentation.

**🌐 Repository of Precomputed Features**
Released a web-based [repository of precomputed input features](https://github.com/KosinskiLab/AlphaPulldown?tab=readme-ov-file#features-database) for multiple model organisms, reducing redundant computations and accelerating workflows.

---

1) Update biopython to the latest version
2) Instructions on how to install cpp4 within the analysis container
3) Fix run_multimer_jobs.py relative paths
4) Remove run_alphafold.py and stereo_chemicsl_props.py from installation
5) Made compatible with python>3.10
6) Multiple minor code optimizations

There is only a single fix: generating features with mmseqs2 previously did not work because of 399
Many thanks to [kashyapchhatbar](https://github.com/kashyapchhatbar) for finding this out and reporting!

1. Merged updated manual, credits to gilep
2. Fixed bug: predictions for homo-oligomers using whole ranges.
3. `ranking_debug.json`: save iptm for multimers OR ptm for monomers.
4. Added `modelcif` dependency to the `pulldown.docker` file and github actions.
5. Used system.entities instead of system.target_entities for `convert_to_modelcif.py`.
6. Added tests for `modelcif` and `run_multimer_jobs.py`.
7. Used the correct docker image from `kosinskilab` for the analysis pipeline.