Risk prediction in early phases of clinical trial protocol design

Perritaz, Alex

Back

Master thesis

Risk prediction in early phases of clinical trial protocol design

Perritaz, Alex
Teodoro, Douglas (Degree supervisor)

2020

76 p.

Mémoire de master: Haute école de gestion de Genève, 2020

English The purpose of this study is to develop a solution able to determine the validity of a specific protocol before moving to a new stage of the clinical trial. This makes it possible to rule out ineffective protocols and to preserve financial resources. While the risks associated with the clinical trial design lead to substantial savings, they also prevent effective and promising drugs from reaching the market due to poor protocol design. This study aims to explore large corpora of clinical trials to learn through the three different types of models: classical machine learning methods, deep learning-based and contextualized language models. A clinical trial collection from clinicaltrials.gov was used to train and assess the models. Risk factors were associated based on the successful termination of clinical trial. The results of classical machine models manage to correctly classify the risks in 43% of cases. The second approach, using deep learning models, obtains an accuracy of 42%. The third approach, on the other hand, using Transformers contributes to a classification up to 63% accurate for the identification of risks.

Language

English

Classification

Information, communication and media sciences

Notes

Haute école de gestion Genève
Information documentaire
hesso:hegge

License

License undefined

Identifiers

RERO DOC 329737

Persistent URL

https://sonar.ch/hesso/documents/314910

Statistics

Document views: 202 File downloads:

Mémoire: 368

SONAR|HES-SO

Master thesis

Risk prediction in early phases of clinical trial protocol design

clinical trials

natural language processing

language models

deep learning

transformers

LSTM

Seq2Seq

word embeddings

Statistics