Softwareprojekt: Machine Learning in the Life Sciences
Katharina Baum
Comments
Throughout this project, we will train various machine learning (ML) methods and evaluate their outcomes. To achieve this, we will prepare and analyze different datasets from the life sciences field, ... read more
Throughout this project, we will train various machine learning (ML) methods and evaluate their outcomes. To achieve this, we will prepare and analyze different datasets from the life sciences field, some of which may be large, for machine learning purposes aligned with predefined questions. These questions can either be provided by us, closely tied to our research interests, or developed collaboratively. A specific application could be in personalized medicine, such as predicting the efficacy of cancer drugs using comprehensive data from cancer cells, or making temporal forecasts, for instance, predicting infection rates during epidemics. We will use Python as the programming language, and we plan to employ modern Python modules for ML, such as scikit-learn, TensorFlow, or PyTorch. Proficiency in Python is a prerequisite. The objective is to create a Python package that delivers reusable code tailored for the specific use case, encompassing preprocessing, training ML models, evaluating results, and documentation (e.g., using sphinx). This software project will run concurrently with the semester.
Update 22.04.: We still have spots available!
If you're interested and want to register in the CM, please contact us at pascal.iversen@fu-berlin.de and pauline.hiort@fu-berlin.de.
We plan to offer one of two possible projects:
(1) Active learning for drug response prediction in cancer: The project aims to improve the prediction of drug efficacy in cancer using deep learning models through active learning.
(2) Deep learning for drug combination response prediction: Here, you will implement a deep learning baseline for predicting drug combinations and compare it with a published method. You will then test the methods with randomized input.
On Whiteboard, you'll find more information about it.
