• Kedar Patwardhan, Kedar Patwardhan, AstraZeneca

At AstraZeneca Oncology Data Science, we committed to unlock the potential of AI/ML-driven data science. Here we demonstrate that pre-treatment clinical lab values and non-invasive CT imaging features can be used to model survival risk in the mNSCLC setting. This is an important step towards improving patient access to Immunotherapy.

• Felipe Pérez-Jvostov

National data infrastructure is a critical enabler of Canada’s genomic research and community-driven collaboration. The success of such infrastructure is dependent on its capacity to address growing demands for data and its ability to enable a diverse range of resource-intensive computational activities. This presentation will focus on the challenges and opportunities of establishing such national data infrastructure in the Canadian context, and highlight the importance of data interoperability across domains and data types to fuel research and innovation in genomic science and beyond.

• Gordon Hanna, Gordon Hanna, Imperial College London
• Tarun Khanna, Tarun Khanna, Imperial College London
• Cecilia Pennica, Cecilia Pennica, Imperial College London
• Suhail Islam, Suhail Islam, Imperial College London
• Michael Sternberg, Michael Sternberg, Imperial College London
• Alessia David, Alessia David, Imperial College London

Missense3D (http://missense3d.bc.ic.ac.uk/) predicts the impact of missense variants on protein structure and reports their structural impact e.g. burial of charged residues. A user can assess the impact of a variant on a monomeric structure, including its transmembrane region, or on a protein complex. Missense3D accepts any structure, including AlphaFold models.

• Suzanne Paley

An overview of the BioCyc website and Pathway Tools software suite, which features an extensive array of capabilities covering genome informatics, pathway informatics, regulatory informatics, and omics data analysis. Several new capabilities will be demonstrated, including multi-omics visualization tools, a new genome browser, and the comparative genome dashboard.

• Nicola Bordin

CATH, now up-to-date with the Protein Data Bank, created with the group of David Jones at UCL the TED resource, classifying over 200m domains from AFDBv4 within the CATH classification and identified over 7k novel folds. TED offers community access via a dedicated web resource, facilitating data visualization and downloads.

• Paulina Dragan
• Dorota Latek
• Przemyslaw Osial

The number of GPCR structures in PDB and their active ligands has
recently become sufficient to apply machine learning in the compound
activity recommendation systems for drug design. GPCRVS [1] is an
efficient machine learning system [2, 3, 4] for the online assessment of
the compound activity against several GPCR targets, including peptide
and protein-binding GPCRs, the most difficult for virtual screening [3].
GPCRVS evaluates compounds in terms of their activity range,
pharmacological effects, and binding modes. GPCRVS evaluates compounds
ranging from classical small molecules to short peptides. Results of
activity class assignment and binding affinity prediction are provided
in comparison with known active ligands of each GPCR receptor type. A
multi-class classification in GPCRVS, handling incomplete and fuzzy
biological data, was validated on ChEMBL-retrieved training data sets
for class B GPCRs and chemokine CC and CXC receptors. Acknowledgments:
National Science Centre in Poland (2020/39/B/NZ2/00584).

Availability: https://gpcrvs.chem.uw.edu.pl

References:

[1] D. Latek, K. Prajapati, M. Merski, P. Dragan, P. Osial. GPCRVS – a
machine learning system for GPCR drug discovery, submitted.

[2] P. Dragan, K. Joshi, A. Atzei, D. Latek Keras/TensorFlow in Drug
Design for Immunity Disorders. Int. J. Mol. Sci. 2023, 24, 15009.

[3] P. Dragan, M. Merski, S. Wisniewski, S.G. Sanmukh, D. Latek
Chemokine Receptors - Structure-Based Virtual Screening Assisted by
Machine Learning. Pharmaceutics 2023, 15(2), 516.

[4] M. Mizera, D. Latek. Ligand-receptor interactions and machine
learning in GCGR and GLP-1R drug discovery. Int. J. Mol. Sci. 2021,
22(8), 4060.

Author: Dorota Latek

• Martin Miller, Martin Miller, AstraZeneca

At AstraZeneca’s Oncology Data Science, we committed to unlock the potential of AI/ML-driven data science. Here, we model clinical endpoints together with >10.000 of DNA and RNA profiled tumour samples from patients progressing on immune checkpoint blockade (ICB). We uncover that the post-ICB tumour microenvironment is fundamentally different in acquired vs primary resistance. At AstraZeneca’s Oncology Data Science, we committed to unlock the potential of AI/ML-driven data science. Here, we model clinical endpoints together with >10.000 of DNA and RNA profiled tumour samples from patients progressing on immune checkpoint blockade (ICB). We uncover that the post-ICB tumour microenvironment is fundamentally different in acquired vs primary resistance.

• Sonika Tyagi

DNA is the blueprint defining all living organisms. Therefore, understanding the nature and
function of DNA is at the core of all biological studies. Rapid advances in DNA sequencing
and computing technologies over the past few decades resulted in large quantities of DNA
generated for diverse experiments, exceeding the growth of all major social media platforms
and astronomy data combined. However, biological data is both complex and
high-dimensional, and is difficult to analyse with conventional methods.
Machine learning is naturally well suited to problems with a large volume of data and
complexity. In particular, applying Natural Language Processing to the genome is
intuitive, since DNA is a natural language. Unique challenges exist in Genome-NLP over
natural languages, including the difficulty of word segmentation or corpus comparison.
To tackle these challenges, we developed the first automated and open-source genomeNLP
workflow that enables efficient and accurate knowledge extraction on biological data,
automating and abstracting preprocessing steps unique to biology. This lowers the barrier to
perform knowledge extraction by both machine learning practitioners and computational
biologists.

• Filippo Utro, Filippo Utro, IBM Research

In the recent years, foundation models (FM) and quantum computing (QC) in healthcare and life science have sparked significant interests. This talk explores the latest effort of IBM Research in FM and QC aiming to accelerate discovery in healthcare and life science. I will delve into 3 different FMs that we are developing to accelerate drug discovery and in different efforts on QC in particular Quantum Machine Learning as a powerful tool discussing some of its application spanning from genomics to diagnostics in medical research. We also will discuss technical challenges, envisioning the new era of FM and QC in healthcare and life science.

• Eli Pollock