Date: 6/12/2020
Attendees: Bharath, Peter, Angelica, Taran, Nathan
Summary: We had a few new attendees on the call so we started with a round of introductions. Bharath is an open source developer who works on DeepChem whenever he finds the time. Peter is an open source developer who works on DeepChem, OpenMM, and Folding at Home. Angelica is a scientist who works at Nurix. Taran is an undergraduate at PES university in Bangalore. Nathan is a graduate student at UPenn.
Bharath has been more a user of DeepChem than a developer this week. From the user experience, the new docs are a major step up for reference. However, the docs still lack examples. Numpy and Scipy’s docs have many user examples scattered throughout. Our docs could benefit from more code samples scattered throughout. He also found a bug in one of the metrics. On the code from, Bharath has been working on getting the docking PR merged in. He hopes to have that ready to merge in within the next few days.
Peter has been working on OpenMM this week and didn’t have time to work on DeepChem issues.
Since the other folks on this call are new, we did a brief round of discussing what drew them to DeepChem and what they were interested in working on.
Angelica is new to using DeepChem, but has looked a bit at the random forest and splitter usage. She’s interested in using more of the library in practice.
Taran found DeepChem online and worked through the tutorials. He’s interested in contributing to DeepChem and has been working on an improvement to the image data augmentation. Bharath suggested that for early feedback, it might be useful to put up a WIP (work in progress) pull request for review. Taran said he’d send over a PR for review in a bit.
Nathan found DeepChem through a talk and is interested in learning more about deep learning applications for the life sciences and chemistry. Nathan’s developed on pymatgen. Bharath mentioned that improved pymatgen integration and the addition of more materials science datasets to MoleculeNet would be a very welcome contribution.
Bharath mentioned that Seyone and Daiki sent notes that they couldn’t make the call this week. Daiki sent an update about his Google Summer of Code work. He’s improved the install time on the tutorials to 2 minutes, and has cleaned up the DeepChem setup code as well. Daiki’s also made progress on implementing GCNs in Jax, and has a base implementation which achieves reasonable accuracy. Daiki is going to work on additional testing and Haiku integration to clean up the code.
As a quick reminder to anyone reading along, the DeepChem developer calls are open to the public! If you’re interested in attending, please send an email to X.Y@gmail.com, where X=bharath, Y=ramsundar.