Logo Logo
Help
Contact
Switch language to German
Diagnosing and predicting clinical outcomes based on computational methods for immune microenvironment patterns: two examples
Diagnosing and predicting clinical outcomes based on computational methods for immune microenvironment patterns: two examples
The immune system is a critical component of the delicate balance between human health and disease, particularly as immunotherapy gains popularity. In order to identify patients at an early stage and develop individualized disease prevention strategies, it is important to systematically and accurately describe the immune environment before disease onset. However, the immune environment is a complex system consisting of immune cells, antibodies, complement, and cytokines. Traditional methods of monitoring immune patterns in clinical settings are limited in accuracy and reliability, which have created a pressing need for more advanced technologies. Therefore, bioinformatics has become an important tool in the field of disease immunology research. In recent years, computational methods and approaches have been developed, which specifically enumerate the immune microenvironment and allow for the further quantification of the complex immune system. The objective of this project is to explore the role of bioinformatics in the prediction of the diagnosis or prognosis of immune-related diseases. The focus of this dissertation is on two types of diseases, each with a different prediction model, which are described separately in two independent chapters due to their respective specificities. In Chapter 2, the immune cell compositions of blood samples from patients with Kawasaki disease (KD)—an immune-mediated inflammation in children—were enumerated. A novel algorithm was developed for predicting KD diagnosis based on this enumeration. Using the model, patients with KD and febrile controls could be well distinguished in the test set, with an AUC of 0.80. In Chapter 3, a study concerning a tumor disease, uveal melanoma (UVM), was conducted. The study integrates the patterns of basement membrane and immunogenic cell death to investigate the immune microenvironment patterns in UVM patients. On this basis, three models using different algorithms were constructed and the optimal one was selected after comparing them in validation set, which was the model generated by the IPF-LASSO algorithm, with an AUC of 0.740, 0.841 and 0.835 for 1-, 3-, and 5-year overall survival, respectively. Furthermore, we assessed its performance on the test sets with different survival outcomes and preliminary investigated its association with the response to UVM immunotherapy. As such, this dissertation highlights how the integration of machine learning and high-throughput data can improve the characterization of the disease-associated immune microenvironment and the development of better prediction models. The study results demonstrate that the bioinformatics-based approach presented in this project holds great potential for predicting the diagnosis or prognosis of immune-related diseases, and could ultimately improve patient outcomes.
Not available
Du, Shangming
2023
English
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Du, Shangming (2023): Diagnosing and predicting clinical outcomes based on computational methods for immune microenvironment patterns: two examples. Dissertation, LMU München: Faculty of Medicine
[thumbnail of Du_Shangming.pdf]
Preview
PDF
Du_Shangming.pdf

7MB

Abstract

The immune system is a critical component of the delicate balance between human health and disease, particularly as immunotherapy gains popularity. In order to identify patients at an early stage and develop individualized disease prevention strategies, it is important to systematically and accurately describe the immune environment before disease onset. However, the immune environment is a complex system consisting of immune cells, antibodies, complement, and cytokines. Traditional methods of monitoring immune patterns in clinical settings are limited in accuracy and reliability, which have created a pressing need for more advanced technologies. Therefore, bioinformatics has become an important tool in the field of disease immunology research. In recent years, computational methods and approaches have been developed, which specifically enumerate the immune microenvironment and allow for the further quantification of the complex immune system. The objective of this project is to explore the role of bioinformatics in the prediction of the diagnosis or prognosis of immune-related diseases. The focus of this dissertation is on two types of diseases, each with a different prediction model, which are described separately in two independent chapters due to their respective specificities. In Chapter 2, the immune cell compositions of blood samples from patients with Kawasaki disease (KD)—an immune-mediated inflammation in children—were enumerated. A novel algorithm was developed for predicting KD diagnosis based on this enumeration. Using the model, patients with KD and febrile controls could be well distinguished in the test set, with an AUC of 0.80. In Chapter 3, a study concerning a tumor disease, uveal melanoma (UVM), was conducted. The study integrates the patterns of basement membrane and immunogenic cell death to investigate the immune microenvironment patterns in UVM patients. On this basis, three models using different algorithms were constructed and the optimal one was selected after comparing them in validation set, which was the model generated by the IPF-LASSO algorithm, with an AUC of 0.740, 0.841 and 0.835 for 1-, 3-, and 5-year overall survival, respectively. Furthermore, we assessed its performance on the test sets with different survival outcomes and preliminary investigated its association with the response to UVM immunotherapy. As such, this dissertation highlights how the integration of machine learning and high-throughput data can improve the characterization of the disease-associated immune microenvironment and the development of better prediction models. The study results demonstrate that the bioinformatics-based approach presented in this project holds great potential for predicting the diagnosis or prognosis of immune-related diseases, and could ultimately improve patient outcomes.