Machine Learning Approach for Personalized Prediction of Renal Graft Survival
| Project/Area Number |
23K17002
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 62010:Life, health and medical informatics-related
|
|---|
| Research Institution | Shonan Kamakura General Hospital, Medical Corporation Tokushukai (Center for Clinical and |
|---|
Principal Investigator |
Salybekov Amankeldi 医療法人徳洲会湘南鎌倉総合病院(臨床研究センター), 湘南先端医学研究所 再生医療開発研究部, 主席研究員 (80850776)
|
|---|
| Project Period (FY) |
2023-04-01 – 2025-03-31
|
| Project Status |
Granted (Fiscal Year 2023)
|
| Budget Amount *help |
¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2024: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
Fiscal Year 2023: ¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
|
|---|
| Keywords | Machine learning / Graft failure prediction / Neural network / Kidney transplantation / Kidney / Transplantation |
|---|
| Outline of Research at the Start |
1) Data preprocessing
2) Phenotype data normalization (histology, laboratory, and clinical variables)
3) Missing data imputation and synthetic data generation
4) Model development ( random forest, LGBM, XgBoost, SVM etc. Important features development.
5) Validation of the dataset and deployment of final model
|
| Outline of Annual Research Achievements |
We developed algorithms for predicting the survival of living and deceased donor grafts before transplantation. We trained a large dataset (n = 135000 for deceased donors and n = 110000 for living donors) obtained from USA & Japanese cohorts. We split training and testing groups into 70/25% and 5% for external validation. All of the missing values of variables were filled up using MisRanger. Then we used LGBM, Random Forest, and reccurrent neural network algorithms. The top 20 predicted variables are further used for graft failure prediction. AUROC for 5-year graft failure prediction were AUROC-97 for living and AUROC-96 for deseased transplants. We also checked the F-1 score. Currently, we are preparing full paper for publication soon. The results were presented at ASN 2023 and JSN 2023.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
Currently, we finalizing obtained data for full paper publication.
|
| Strategy for Future Research Activity |
We plan to develop software for prediction of graft failure. Currently we are working on it.
|
Report
(1 results)
Research Products
(5 results)
