Machine learning in personalized medicine : modeling multiple myeloma treatment responses
Uramoto, Leo (2020-10-23)
Machine learning in personalized medicine : modeling multiple myeloma treatment responses
(23.10.2020)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
avoin
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2020110689481
https://urn.fi/URN:NBN:fi-fe2020110689481
Tiivistelmä
This thesis explores various machine learning models and attempts to use them to gain insight on individual differences in treatment responses of patients diagnosed with multiple myeloma, a hematopic malignancy representing approximately 1% of all cancers. While the 5-year survival rate exceeds 50%, treatments are not personalized and treatment results are heterogeneous.
Machine learning systems processing large amounts of patient data could help tailor the treatments to individual patients. Two research goals are set; understanding the medical trajectory of a patient by predicting the best treatment response they will reach in the future and predicting how long a given patient can remain on a given drug.
The work is split into two main parts. In the first part the mathematical concepts of machine learning and several commonly used machine learning models are introduced. In the second part the theoretical portions are applied by training the introduced models on a data set from the CoMMpass study by the Multiple Myeloma Research Foundation to answer the research questions.
The results from the medical trajectory prediction are somewhat promising and with additional research could become accurate enough for real-world use. The drug duration prediction task turns out too complicated for the limited methodology of this thesis. A discussion of the results and possible improvements on the methods are provided.
Machine learning systems processing large amounts of patient data could help tailor the treatments to individual patients. Two research goals are set; understanding the medical trajectory of a patient by predicting the best treatment response they will reach in the future and predicting how long a given patient can remain on a given drug.
The work is split into two main parts. In the first part the mathematical concepts of machine learning and several commonly used machine learning models are introduced. In the second part the theoretical portions are applied by training the introduced models on a data set from the CoMMpass study by the Multiple Myeloma Research Foundation to answer the research questions.
The results from the medical trajectory prediction are somewhat promising and with additional research could become accurate enough for real-world use. The drug duration prediction task turns out too complicated for the limited methodology of this thesis. A discussion of the results and possible improvements on the methods are provided.