Leo Klarner

PhD Student in AI for Drug Discovery at the University of Oxford


 
website_pic.jpg

leo.klarner [at] stats.ox.ac.uk


I’m a final-year PhD student in the Oxford Protein Informatics (OPIG) and CompStats & Machine Learning (OxCSML) groups. My research focuses on developing more robust and data-efficient generative models for applications in early-stage drug discovery and protein design. Specifically, I’m working on improving out-of-distribution generalisation (ICML 2023, ICML 2024) and generative modelling under domain-informed constraints (TMLR, 2023, NeurIPS 2023).

This work requires familiarity with both modern deep learning and the realities of practical drug discovery, which I have been fortunate to pick up from my supervisors Yee Whye Teh (Oxford/DeepMind), Charlotte Deane (Oxford) and Garrett Morris (Oxford), as well as Torsten Schindler and Michael Reutlinger (Roche). My work is funded by a Clarendon Scholarship (Oxford’s flagship academic merit scholarship for graduate students) and additional partnership awards from Brasenose College, Oxford and Roche.

I feel strongly about understanding the data I work with and have spent many weeks in ChEMBL, PubChem and the PDB. I am also interested in curating more realistic datasets and benchmarks (ICML 2023, AI for Chemistry Best Poster Award) and am a main contributor of the open-source Gauche package (NeurIPS 2023).

Before starting my PhD, I completed a BSc. in Interdisciplinary Sciences (chemistry, biology and CS) at ETH Zürich. During this time, I had the opportunity to design and synthesise antimicrobial peptides with Prof Gisbert Schneider and engineer bacteria for targeted cancer therapy with Prof Simone Schürle-Finke.

Publications

2024

  1. Context-Guided Diffusion for Out-of-Distribution Molecular and Protein Design
    Leo Klarner, Tim G. J. Rudner, Garrett M Morris, Charlotte Deane, and Yee Whye Teh
    International Conference on Machine Learning (ICML), 2024

2023

  1. Metropolis Sampling for Constrained Diffusion Models
    Nic Fishman, Leo Klarner, Emile Mathieu, Michael Hutchinson, and Valentin De Bortoli
    Advances in Neural Information Processing Systems (NeurIPS), 2023
  2. GAUCHE: A Library for Gaussian Processes in Chemistry
    Ryan-Rhys Griffiths, Leo Klarner, Henry Moss, Aditya Ravuri, Sang T. Truong, Yuanqi Du, Samuel Don Stanton, Gary Tom, Bojana Ranković, Arian Rokkum Jamasb ... Alpha Lee, Bingqing Cheng, Alan Aspuru-Guzik, Philippe Schwaller, and Jian Tang
    Advances in Neural Information Processing Systems (NeurIPS), 2023
  3. klarner2023qsavi.png
    Drug Discovery under Covariate Shift with Domain-Informed Prior Distributions over Functions
    Leo Klarner, Tim G. J. Rudner, Michael Reutlinger, Torsten Schindler, Garrett M. Morris, Charlotte Deane, and Yee Whye Teh
    International Conference on Machine Learning (ICML), 2023
  4. Diffusion Models for Constrained Domains
    Nic Fishman, Leo Klarner, Valentin De Bortoli, Emile Mathieu, and Michael Hutchinson
    Transactions on Machine Learning Research (TMLR), 2023

2022

  1. Bias in the Benchmark: Systematic experimental errors in bioactivity databases confound
    multi-task and meta-learning algorithms
    Leo Klarner, Michael Reutlinger, Torsten Schindler, Charlotte Deane, and Garrett Morris
    2nd ICML AI for Science Workshop, 2022
    Best Poster Award at 5th AI for Chemistry Conference, 2022