cheminformania.com
I like chemistry, computers and drug discovery. Sometimes blogs about cheminformatics, ML and neural networks for chemistry at cheminformania.com. Open source fanatic and science preprinter.
54 posts
409 followers
127 following
Regular Contributor
Active Commenter
comment in response to
post
📢 We’d Love Your Feedback!
Check out the new docs and let us know:
• Are they helping your workflows?
• How are you using Scikit-Mol to build chemistry models?
Your insights shape the future of Scikit-Mol! 💬
👉 scikit-mol.readthedocs.io/en/latest/
comment in response to
post
🧪 What’s New in Scikit-Mol?
1️⃣ Comprehensive API & Notebook Docs
• Clear, structured documentation for both the API and example notebooks via MkDocs.
2️⃣ Automated Builds
• Docs are now auto-generated & hosted on Read the Docs—always up-to-date!
#MachineLearning #Cheminformatics
comment in response to
post
🔬 What is Scikit-Mol?
Scikit-Mol is an open-source library that bridges RDKit featurization with machine learning.
It provides Scikit-Learn compatible classes for molecular fingerprints and descriptors—simplifying chemistry + ML workflows! 🧪
#OpenSource #Scikit-Learn #RDkit
comment in response to
post
Oh. now its there😄
comment in response to
post
If you make some cool custom links, be sure to reach out. It could use a bit of expansion with more links. Is it reaction transforms or what are you working on?
comment in response to
post
Somehow bluesky eats the link, but its easy to google
comment in response to
post
github.com/EBjerrum/pdc...
comment in response to
post
I made pdchemchain for reusable and easy dataprocessing of dataframes with a chemical view. I strived to make it easy to make new reusable classes while still getting all the goodies of self documentation and serialization for command line reusage. Take a look, and let me know if we should chat.
comment in response to
post
6/6
💡 We’d Love Your Feedback!
Are you using Scikit-Mol in your projects? Got ideas, feature requests, or want to contribute?
Let’s make Scikit-Mol even better—together. 🙌
#Cheminformatics #MachineLearning #OpenSource #Python #RDKit #ScikitLearn #ScikitMol
comment in response to
post
5/6
4️⃣ Deprecations & Build System Updates:
• Python 3.8 is now deprecated.
• Switched to uv & hatchling for build/dependency management (see CONTRIBUTING.md).
• Added plotting.py for benchmarks + updated Notebook 7 with new benchmarks.
comment in response to
post
4/6
2️⃣ API Update – Consistency with scikit-learn:
• parallel → n_jobs (matches scikit-learn API)
• If you used parallel=True, switch to n_jobs=-1 for full CPU usage.
3️⃣ Bug Fix:
Fixed a potential issue with safe_inference in parallel runs.
#ScikitLearn #RDKit
comment in response to
post
3/6
⚡ What’s New in Scikit-Mol 0.4.6?
1️⃣ Faster Parallel Processing:
Migrated from multiprocessing to joblib, fixing performance issues on Windows & Mac.
✅ Test times down from ~7 mins to ~1.5 mins on Windows!
✅ More robust, easier to maintain code.
comment in response to
post
2/6
🔬 What is Scikit-Mol?
An open-source library that bridges #RDKit featurization with ML. With @scikit-learn.org compatible classes for molecular fingerprints & descriptors, it streamlines chemistry + ML workflows.
Learn More: github.com/EBjerrum/sci...
comment in response to
post
Ambitous, but your efforts are really appreciated.
comment in response to
post
Yes!
comment in response to
post
Open-source collaboration drives innovation! By sharing RDEditor, we gained valuable contributions that make it better for everyone.
Check it out:
📦 pypi.org/project/rded...
📂 github.com/EBjerrum/rde...
Let’s keep building together!
comment in response to
post
🛠 Bug Fixes:
• Stabilized bond lengths
• Resolved stereo rendering issues during coordinate recalculations.
comment in response to
post
🎨 New Features:
• Stereoinformation Pre-Drawing: Fixed before rendering, with an option to switch off for raw molecular graph display.
• Autoscaling: Molecule visuals now auto-scale for a cleaner, more intuitive drawing experience.
comment in response to
post
🔍 Wiggly Bonds for Undefined Stereo:
You can now display wiggly bonds at undefined stereocenters for better representation of ambiguous stereochemistry. Customization options available in the new drawing settings.
comment in response to
post
Post 5:
📢 How Are You Using Scikit-Mol?
We’d love to hear how it’s helping your projects! Are there features you’re missing or ideas you’d like to see implemented? Drop us a line and share your thoughts. #OpenSource #Cheminformatics
comment in response to
post
💡 Details from the PRs:
• Better handling of invalid SMILES and more robust SafeInferenceWrapper.
• Python 3.13 support, a new linter/formatter, and even a PyPI-visible logo! 🎨
Huge shoutout to Anton Siomchen for the excellent work! 🙌
comment in response to
post
🛠️ What’s New in v0.4.5?
1️⃣ Bug fixes:
• Safe-inference now handles batches with invalid molecules seamlessly.
• Parallel execution fixed for small sample sizes.
2️⃣ Under-the-hood improvements:
• Automated formatting, pre-commit hooks, badges, and README fixes.
comment in response to
post
🔬 What’s Scikit-Mol?
It’s an open-source library that bridges #RDKit featurization with ML models. Think #Scikit-Learn -compatible classes for fingerprints and descriptors. Easy, efficient, and chemistry-savvy! 🌟
comment in response to
post
This comes to mind. youtu.be/u1cTRn5tBs8?...
Trump/Elmo are likely wanting Kvanefjeld mining.
comment in response to
post
A Thank You
To everyone who’s supported me in 2024—thank you. Challenges turned into opportunities because of your encouragement and collaboration. Let’s make 2025 a year of impact and innovation.
Happy New Year! 🎉
#Grateful #NewYear
comment in response to
post
Moving Forward
2025 is on the horizon, and I’m optimistic about new opportunities.
If you know of roles or connections in drug discovery, ML, AI, or open-source development, I’d love to connect! Let’s keep building together and advancing science.
#Collaboration #Opportunities
comment in response to
post
Open-Source Highlights
Grateful for the incredible collaborations in open-source science this year!
🌟 Community contributions have strengthened Scikit-Mol & rdEditor.
💡 New projects: pdChemChain & RandomNets.
Together, we’re making tools more accessible and impactful.
#OpenSource #Cheminformatics
comment in response to
post
Thanks, looks interesting
comment in response to
post
Timely release then :-) Let me know how it goes
comment in response to
post
🎁 Tried RandomNets yet?
Ensemble ML done right:
🔹 25% avg boost in R² & MSE across 133 datasets
🔹 Uncertainty via std. dev. in predictions
🔹 Simple PyTorch Lightning workflow
GitHub: github.com/EBjerrum/Ran...
comment in response to
post
Why train networks in slow loops? 🚀 RandomNets combines vectorized feature masking + ensembles for fast training/inference in PyTorch Lightning. Better predictions, faster training, built-in uncertainty.
Tutorial: github.com/EBjerrum/Ran...
comment in response to
post
For you? yes
comment in response to
post
Yes, sometimes its more intense, but I try to at least keep the streak going😆
comment in response to
post
Better to teach free alternatives than proprietary software. However, then the teachers need to learn first
comment in response to
post
Sure, its a extra incentice. Anyway, you really crush it with Duo, impressive. I think I found you in the app.
comment in response to
post
I feel for you. But is the league and gamification not just tricks to get to learn some language?
comment in response to
post
Oh, I thought it was an activity cliff