The Data Science and Learning Division at the Argonne National Laboratory has an opening for a postdoctoral researcher to develop computational methodologies for comparing deep learning drug response models to advance precision oncology. This is a multi-year project in collaboration with the Frederick National Laboratory for Cancer Research (FNLCR) and makes use of large GPU clusters, advanced AI accelerators, and DOE leadership class supercomputers.

The successful candidate will work in an interdisciplinary team to develop a computational framework for comparing and improving deep learning models for drug response prediction. Prediction of cancer treatment response using machine learning algorithms holds great promise for the future of precision oncology. While various deep learning techniques have been explored for this purpose, the lack of a rigorous framework for comparing models and evaluating their potential utility in clinical practice impedes the adoption of these methods in mainstream cancer research. By leveraging the knowledge of deep learning and cancer biology you will collaborate with scientists in developing a framework for large-scale analyses of deep learning models. You will develop workflows that will ultimately enable personalized recommendations of treatments, drug repurposing, and new drug development. The successful candidate will be an integral part of a large group of computational scientists in the Data Science and Learning division.

Responsibilities:

• Develop methodologies to assess the generalizability, reproducibility, performance, and robustness of deep learning models for drug response prediction
• Propose novel data-driven approaches to improve existing drug response prediction models to advance the personalized treatment of cancer
• Implement large-scale workflows for deep learning, including hyperparameter optimization (HPO), neural architecture search (NAS), and distributed model training in high-performance computing (HPC) environments
• Establish benchmark datasets and evaluation criteria to assess the performance of deep learning models for precision oncology
• Participate in hackathons with project collaborators and the broader community of deep learning developers and cancer researchers
• Publish and present findings based on comparisons of community-developed drug response prediction models

Position Requirements

• Recent or soon-to-be completed Ph.D. (typically within 0-3 years) in computer science, electrical engineering, statistics, bioinformatics, computational biology, or related fields
• Comprehensive experience implementing and evaluating deep learning models that use open-source frameworks such as TensorFlow (Keras), PyTorch, or MXNet (Gluon)
• Knowledge of deep learning methods such as transfer learning, multi-task learning, and auto encoders
• Experience with Python machine learning modules such as scikit-learn, numpy and pandas
• Experience with version control including Git and GitHub
• Ability to communicate research outcomes through presentations and publications, present findings at conferences, and participate in the development of research proposals
• Ability to model Argonne’s Core Values: Impact, Safety, Respect, Integrity, and Teamwork
• Experience applying statistical analysis and machine learning for cancer research
• Experience installing deep learning frameworks in a Linux operating system environment
• Experience building containers with singularity and/or docker
• Experience with data analysis and programming in high-performance computing (HPC) environment
• Experience curating and analyzing cancer multi-omics data (e.g., genomics, transcriptomics)

Job Family

Job Profile

Worker Type

Time Type