Some Notes From the 2018 RDKit UGM

Last week I had the pleasure of attending the RDKit User Group meeting in Cambridge, UK.  This was my first RDKit UGM, and it was great.  I had the opportunity to catch up with a lot of people I hadn’t seen for a while and learned about a lot of exciting Open Source Cheminformatics. In this post, I’ve tried to summarize some of what took place and to present some links to relevant software and literature.  This won’t be a complete recitation of everything that took place, but hopefully, it will provide an overview for those who’d like to dig deeper.  I’ll link the slide decks as they become available.  Please let me know if I’ve missed or misinterpreted anything.


Slides from the meeting are available in GitHub https://github.com/rdkit/UGM_2018

Wednesday, September 19th

Greg Landrum, KNIME/T5 Informatics, Welcome and Intro (slides)
Greg provided a bit of history of the RDKit as well as an intro to some of the newer features.
  • C++ code has been modernized to C++ 14, greatly simplifying things like iteration
  • New sensible defaults for many functions
  • Coordinate generation code from Schrödinger for prettier depictions
  • Code to depict fingerprint bits
  • A JSON-based format for interchange between programs
  • New 3D descriptors
  • SVG rendering with chemical metadata
Pat Lorton, Schrödinger, Similarity searching a billion compounds in real time 
Pat showed a very nice open source implementation of fingerprint similarity searching on a GPU.
Sereina Riniker, ETH, News from 3D (new features of the conformer generator)
Sereina Described some enhancements to Experimental Torsion Knowledge Distance Geometry (ETKDG) conformer generation method
Alpha Lee, Cambridge University, Uncertainty quantification with Bayesian molecular deep learning (slides)
Alpha discussed ways of dealing with uncertainty in molecular deep learning
Paulo Tosco, Cresset, Lightning Talk
Paulo showed some nice examples of how the RDKit can be seamlessly integrated with Cresset’s Flare toolkit, which is accessible from Python
Tim Dudgeon, Informatics Matters, Lightning Talk
Tim showed a couple of things, the first was Squonk, which appears to be a Jupyter (IPython) Notebook on steroids, definitely worth a look
Tim also described a graph database implementation of a fragment network similar to the one published by Astex
Susan Leung, GSoC RDKit MolVS Integration Project (slides)
Susan discussed her work on a Google Summer of Code project to integrate some of the features from the MolVS virtual screening toolkit into the RDKit
Boran Adas, RDKit's new fingerprint generators (slides)
Boran discussed another Google Summer of Code project designed to unify the many different fingerprint implementations and interfaces in the RDKit under a single consistent framework.  This work should also simplify the addition of new fingerprint types.

Nicholas Firth, Evariste Technologies, Multiparameter optimization using RDKit and scipy: what's the chance of success? (slides)
Nicholas described some of the approaches taken toward quantitative drug design at Evariste Technologies
Marina Fedorova, Novo Nordisk, Computational tools for metabolic pathways prediction (slides)
Marina provided an overview of computational tools being designed to optimize the yield of biosynthetic pathways
  • Building genome-scale models of metabolic pathways
  • Reactions currently encoded as ChemAxon SMIRKS, some interest in translating to Reaction SMARTS
  • Pathways are traced based on Tanimoto similarity to the target compound
Brian Cole, DE Shaw Research, SMILES-driven de novo design engine 
Brian demonstrated how he could hack SMILES strings as a tool for de-novo design of novel molecules.

 Pat Walters, Relay Therapeutics,  A Few (Hopefully) Interesting Open Source Projects Built On The RDKit (slides)
A bunch of stuff that you probably already read about in this blog.

Thursday, September 20th

Joshua Meyers and Matthew Sellwood, BenevolentAI, Rediscovering R-Group Descriptors with RDKit
https://pubs.acs.org/doi/10.1021/ci025589v
Joshua and Matthew described the implementation of R-group descriptors and their subsequent use in identifying bioisosteres
  • The group agreed that there is a need for an open bioisostere database.  Given the availability of software for identifying matched molecular pairs and data available in ChEMBL, it should be possible to collaboratively generate such a database.  
  • Some recent work by the Bajorath group in identifying congeneric series may provide a good starting point. https://www.future-science.com/doi/abs/10.4155/fsoa-2017-0135
Daria Goldmann, KNIME, It's not just for Python! Interacting with chemical data using KNIME and the RDKit (slides)
Daria provided a very nice overview of the many ways that the KNIME platform integrates with the RDKit
Noel O'Boyle, NextMove Software, A de facto standard or a free-for-all? A benchmark for reading SMILES (slides)
Noel described efforts to establish benchmark sets for the parsing and interpretation of SMILES strings
Thomas Evangelidis, CEITEC, Ligand scaffold optimization guided by artificial intelligence
Thomas described his work in combining multiple machine learning methods to build models which
have performed well in a number of recent blind challenges.
Benjamin Tehan + Rob Smith, Heptares, RDKit in the modern Biotech (slides)
Ben and Rob described the many ways that the RDKit is being used at Heptares
Lewis Mervin + Natalia Aniceto, University of Cambridge, In silico protein target prediction with reliability-density neighborhood applicability domain analysis (slides)
Lewis and Natalia described several iterations of the PIDGIN program for target prediction using Random Forests
Marwin Segler, BenevolentAI,  Computer-aided synthesis planning
Marwin described his work on search algorithms for synthesis planning 
Jacob Spiegel, University of Pittsburgh,  Genetic algorithm for de novo computer-aided drug design
Jacob described some recent developments in the AutoGrow program that uses genetic algorithms for de-novo drug design
Andrea Morger, Charité - Universitätsmedizin Berlin, Machine learning and conformal prediction to support in silico toxicology
Andrea described work on using conformal prediction to establish the domain of applicability for predictive models
Lukas Pravda, PDB Europe, The use of RDKit in the Protein Data Bank in Europe to handle small molecules chemistry and display protein-ligand interactions (slides)
Lukas described a number of interesting Python utilities that the European PDB is developing to integrate the processing of proteins and small molecules, many of these tools integrate with the RDKit
Roger Sayle, NextMove Software, Deceptively Simple: How some cheminformatics problems can be more complicated than they appear (slides)
The undisputed king of optimization shows us that few things are as simple as we think they are
  • Calculating molecular weight
  • Counting lines in a text file
  • Determining percentages

A Few Other Interesting Projects

EyeMol: The most interactive molecular dataset manager
https://pikairos.eu/eyemol/

Fast Clustering of Large Datasets
https://github.com/iwatobipen?tab=repositories
https://github.com/fujimizu/bayon

An Open Source molecule editor
Https://github.com/EBjerrum/rdeditor

LiteMol JavaScript PDB viewer
http://webchemdev.ncbr.muni.cz/Litemol/Viewer/


Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Comments

Post a Comment

Popular posts from this blog

AI in Drug Discovery 2023 - A Highly Opinionated Literature Review (Part I)

Image
Here’s the first part of my review of some interesting machine learning (ML) papers I read in 2023.  As with the previous editions , this shouldn’t be considered a comprehensive review.  The papers covered here reflect my research interests and biases, and I’ve certainly overlooked areas that others consider vital.  This post is pretty long, so I've split it into three parts, with parts II and III to be posted in the next couple of weeks.    I. Docking, protein structure prediction, and benchmarking II. Large Language Models, active learning, federated learning, generative models, and explainable AI III. Review articles 2023 was a bit of a mixed bag for AI in drug discovery.  Several groups reported that the deep learning methods for protein-ligand docking weren’t quite what they were initially cracked up to be.  AlphaFold2 became pervasive, and people started to investigate, with mixed success, the utility of predicted protein structures.  There were reports of significant advanc
Read more

Generative Molecular Design Isn't As Easy As People Make It Look

Image
I was taken aback by a recent CNBC article entitled “ Generative AI will be designing new drugs all on its own in the near future ”.  I should know better than to pay attention to AI articles in the popular press, but I feel that even scientists working in drug discovery may have a skewed perception of what generative AI can and can’t do.  To understand exactly what’s involved, it might be instructive to walk through a typical generative molecular design workflow and point out a few things.  First, these programs are far from autonomous.  Even when presented with a well-defined problem, generative algorithms produce a tremendous amount of nonsense.  Second, domain expertise is essential when sifting through the molecules produced by a generative algorithm.  Without a significant medicinal chemistry background, one can’t make sense of the results.  Third, while a few nuggets exist in the generative modeling output, a lot of work and good old-fashioned cheminformatics are required to ext
Read more

AI in Drug Discovery - A Highly Opinionated Literature Review (Part II)

Image
Picking up where we left off in Part I , this post covers several other ML in drug discovery topics that interested me in 2023.  Some areas, like large language models, are new, and most of the work is at the proof-of-concept stage.  Others, like active learning, are more mature, and several groups are starting to explore nuances of the methods.   Here’s the structure of Part II.  4. Large Language Models 5. Active Learning 6. Federated Learning 7. Generative Models 8. Explainable AI 9. Other Stuff 4. Large Language Models The emergence of GPT-4 and ChatGPT brought considerable attention to large language models (LLMs) in 2023.  In November and December, several large pharmas held “AI Day” presentations featuring LLM applications for clinical trial data analysis. Many of these groups demonstrated the ability of LLMs to ingest large bodies of unstructured clinical data and subsequently generate tables and reports based on natural language queries.  Aside from some very brief demos on co
Read more