Tobias Hegelund Olsen

I specialise in applied research, focusing on developing computational tools for the discovery and design of therapeutic proteins, such as antibodies, T-cell receptors (TCRs), and other novel modalities. For this, I make heavy use of new deep learning techniques, but also other data science approaches. The core objective of my work is to integrate computational design pipelines with laboratory workflows, creating a single, highly efficient system for rational design of therapeutics.

I did my PhD in the Oxford Protein Informatics Group (OPIG), University of Oxford, under supervision of Prof. Charlotte Deane, MDE. I have extensive experience with developing new models for immunogenicity, many different developability issues, paratope prediction, structure prediction and affinity prediction. I also have experience collating, processing and working with both structural data and billions of amino acid sequences for both antibodies and TCRs. Additionally, I have worked on predicting useful bioactivities for peptides.

Published and available tools

  • OTS: The Observed T cell receptor Space database enables paired-chain repertoire mining, coherence analysis and language modelling (paper).
  • AbLang-2: Addressing the antibody germline bias and its effect on language models for improved antibody design (paper).
  • AntiFold: Improved antibody structure-based design using inverse folding (paper).
  • KA-Search: A tool for rapid and exhaustive sequence search of billions of antibodies, enabling exploration of the mutational space of antibodies (paper).
  • AbLang: An antibody specific language model for therapeutic antibody engineering (paper).
  • OAS: A diverse database of cleaned, annotated, and translated unpaired and paired antibody sequences (paper).
  • AnOxPePred: Prediction of antioxidative properties of peptides using deep learning (paper).
  • proABC-2: PRediction of AntiBody contacts v2 and its application to information-driven docking. (paper).
  • TCRpMHCmodels: Structural modelling of TCR-pMHC class I complexes (paper).
    • Media

      I can be contacted on LinkedIn and Twitter @HegelundOlsen.

    • GitHub

      Any code I write, and which is public, I will try and make available on my Github.

Experience and Education

  • Investigator, Protein Design and Informatics • GSK, Oct. 2023 - Present

    Applied research for developing computational tools for discovery and design of therapeutic biologics.
  • PhD candidate in Immunoinformatics • University of Oxford, Oct. 2019 - Sep. 2023

    Investigating B-cell repertoire data using deep learning approaches to aid in the development of antibody therapeutics.
  • Research Assistant • Technical University of Denmark, Sep. 2018 - Jun. 2019

    Designing in silico tools for prediction of peptide functions, a deep learning approach for predicting the antibody paratope and de-immunogenization of proteins. Supervised by Professor Paolo Marcatili.
  • M.Sc. Eng. in Pharmaceutical Design and Engineering • Technical University of Denmark, 2016 - 2018

    Thesis: Combining deep learning and structural modelling to predict T cell receptor specificity. Supervised by Professor Paolo Marcatili and Postdoc Kamilla Kjærgaard Jensen at the Department of Bio and Health Informatics, DTU.
  • Student Assistant • Technical University of Denmark, Mar. 2018 - Aug. 2018

    Designing in silico tools for prediction of peptide functions. Supervised by Professor Paolo Marcatili and Professor Egon Bech Hansen.
  • Exchange Program • Seoul National University, South Korea, Sep. 2017 - Dec. 2017

  • B.Sc. in Biotechnology • Technical University of Denmark, 2013 - 2016

    Thesis: Investigation of the Substrate Specificity Determinants of Barley Limit Dextrinase. Supervised by Professor Birte Svensson and Ph.D. Susan Andersen at the Enzyme and Protein Chemistry Group, DTU.
  • Exchange Program • Renssalaer Polytechnic Institute, USA, Sep. 2015 - Dec. 2015

Selected Publications

Contact

Contact me if you have any questions about the tools I have been involved in creating or any other enquiries. The best way is by LinkedIn.

Details

Address
Phone
(000) 000-0000 x 0000
Email