UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Machine Learning Methods for Phenotyping in Drug Discovery

Vauvelle, Andre; (2024) Machine Learning Methods for Phenotyping in Drug Discovery. Doctoral thesis (Ph.D), UCL (University College London). Green open access

[thumbnail of PhD_Thesis_minor_corrections.pdf]
Preview
Text
PhD_Thesis_minor_corrections.pdf - Accepted Version

Download (5MB) | Preview

Abstract

The surge in availability of Electronic Health Record (EHR) data presents both opportunities and challenges for healthcare research. Effective phenotyping, the process of identifying observable traits related to genetic or environmental variations, is integral to understanding diseases and advancing drug discovery. With the increasing complexity of EHR data, traditional methods often fall short. Machine learning, due to its adaptive and data-driven nature, holds the promise of more accurately deciphering patterns within EHR, thus enhancing the precision and scope of phenotyping. Within this work, we address a broad spectrum of challenges associated with structured EHR data and its practical applications, highlighting our ef- forts to enhance phenotyping in the context of drug discovery. The initial research explores EHR’s temporal dynamics, leveraging the signature trans- form as an alternative modelling paradigm for sequential data in heart failure prediction models. Subsequent investigations confront the pervasive issue of label noise in EHR data. By integrating positive and unlabelled learning with transformer architectures, we attempt to enrich cohorts by identifying missed diagnoses and recover genomic associations with greater power. Moreover, the work navigates the realm of patient subtyping, weighing the merits of reconstruction against outcome objectives to forge accurate EHR data repre- sentations. In the final chapter of this research, we advance the field of survival analysis by integrating differentiable sorting methods with partial order super- vision. The method serves as an alternative to the conventional Cox’s partial likelihood with the advantage of a transitive inductive prior.

Type: Thesis (Doctoral)
Qualification: Ph.D
Title: Machine Learning Methods for Phenotyping in Drug Discovery
Open access status: An open access version is available from UCL Discovery
Language: English
Additional information: Copyright © The Author 2024. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/).
UCL classification: UCL
UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences
UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Population Health Sciences > Institute of Health Informatics
URI: https://discovery.ucl.ac.uk/id/eprint/10197859
Downloads since deposit
33Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item