Global Research Postdoctoral Fellow - AI for Patient Stratification and Precision Medicine

Job description

Job Title:

Global Research Postdoctoral Fellow - AI for Patient Stratification and Precision Medicine

Our Team:

The Computational Biology Cluster and the Biomarker & Patient Stratification Cluster are part of the Precision Medicine & Computation Biology (PMCB) global research function at Sanofi. This multidisciplinary group has developed a wealth of biomedical and multi-omics data and is at the frontier of developing advanced AI methodologies that influence the next generation of precision medicine. We seek a motivated postdoctoral fellow to work within the two cluster teams on next generation of foundational models for biology.

Job Description:

As a postdoctoral fellow, you will play a critical role in advancing target discovery and patient stratification methods through the harnessing of deep learning models that integrate multi-modal data.

Your responsibilities will include:

· 
Collaborating closely with interdisciplinary teams to develop and optimize foundational deep learning models for identifying patient subgroups.

· 
Utilizing large datasets including genetic, transcriptomic, single cell, proteomic, imaging and electronic medical record data to train foundational deep learning models on disease specific context.

· 
Contribute to enhancing the precision of patient stratification and disease endotyping.

· 
Contributing to the identification and prioritization of targeted treatments for heterogeneous diseases.

· 
Developing biomarkers that facilitate the classification of patients into identified endotypes.

· 
Participating and leading manuscript preparations and submission, and presentation of findings in international forums internally and externally.

Minimum Required Skills:

· 
Solid experience in deep learning and artificial intelligence methodologies.

· 
Knowledge and experience with foundational/generative AI models for omics or medical record data.

· 
Proficient in machine learning and deep learning frameworks such as TensorFlow or PyTorch.

· 
Proficiency in programming languages such as Python or R.

· 
Solid understanding of computational biology, bioinformatics, or a related field.

· 
Demonstrated ability in data analysis and visualization.

· 
Strong communication skills and ability to work in a collaborative multidisciplinary research environment.

Preferred Skills:

· 
Knowledge of patient stratification and precision medicine concepts.

· 
Experience working with multi-omics data is a plus.

· 
Multi-GPU cloud computing experience

Education:

· 
Ph.D. in Computational Biology, Bioinformatics, Computer Science, or a closely related field.

Project Description:

Patient stratification is a crucial component of precision medicine, as it enables the identification of distinct patient subgroups with different underlying molecular disease profiles, clinical manifestations and response to therapy. Stratification allows for the development of targeted treatments that are more effective, breaking observed efficacy ceilings while reducing adverse events for patients. For heterogenous diseases like Inflammatory Bowel Disease or Parkinson's disease (which has limited disease-modifying treatments available) patient stratification is especially important to identify the right targets for the right patients and thus improve clinical efficacy rates.

With the availability of large amounts of multi-omics and electronic health record (EHR) data deep learning models can leverage an unprecedented wealth of information to identify patient subgroups that were previously unrecognized. However, one limitation of current patient stratification methods is that they typically consider only one layer of data, leaving a lot of value unused. By using deep learning models that can integrate multiple layers of data, we can improve the accuracy and precision of patient stratification.

One challenge is that deep multi-omics and EHR data are rarely available together in the same cohort. Another challenge is that developing complex multi-omics biomarker models can be costly and difficult to translate into a clinical setting. To address these challenges, we propose to investigate deep learning models that can derive patient endotypes from multi-omics and real-world data, in disease specific context. We will consider foundational deep learning models that can be used to infer omics layers when they are not available and identify targets associated with disease progression in specific endotypes. Finally, we will develop simple biomarkers that can be used in a clinical setting to classify patients into the identified endotypes and inform on treatment decisions.

We will explore various foundational models, including autoencoders, recurrent autoencoders, and adversarial neural networks. Foundational autoencoder models have already been developed for single cell multi-omic data integration (Lotfollahi et al.) and several models have been proposed for EHR (Landi et al.), including recurrent autoencoders that can be used to simulate records (Merkelbach et al.). Stanford HAI provides a recent review of existing method in this last domain. We will adapt and combine these approaches to integrate genetic, single cell, bulk transcriptomic, proteomic and EHR data, as well as other data type as relevant. These models will be used for patient clustering and visualization, and developing simpler classification models for clinical biomarkers. Furthermore, we will use these models to infer missing omics and medical record layers and to identify therapeutic targets.

We will use a variety of datasets, including EHR and genetics data from private cohort data, to train pan-disease foundational models and disease specialized models. One of the important aspects of the project is to correct and adapt models across different technologies, populations and healthcare systems. This will enable us to develop models that are robust and applicable across different healthcare settings, thereby facilitating the translation of our findings into clinical practice.

References :

Lotfollahi, Mohammad, Anastasia Litinetskaya, and Fabian J. Theis. "Multigrate: single-cell multi-omic data integration." BioRxiv (2022): 2022-03.

Landi, Isotta, et al. "Deep representation learning of electronic health records to unlock patient stratification at scale." NPJ digital medicine 3.1 (2020): 96.

Merkelbach, Kilian, et al. "Novel architecture for gated recurrent unit autoencoder trained on time series from electronic health records enables detection of ICU patient subgroups." Scientific Reports 13.1 (2023): 4053.

Sanofi Inc. and its U.S. affiliates are Equal Opportunity and Affirmative Action employers committed to a culturally diverse workforce. All qualified applicants will receive consideration for employment without regard to race; color; creed; religion; national origin; age; ancestry; nationality; marital, domestic partnership or civil union status; sex, gender, gender identity or expression; affectional or sexual orientation; disability; veteran or military status or liability for military status; domestic violence victim status; atypical cellular or blood trait; genetic information (including the refusal to submit to genetic testing) or any other characteristic protected by law.

#GD-SA 

At Sanofi diversity and inclusion is foundational to how we operate and embedded in our Core Values. We recognize to truly tap into the richness diversity brings we must lead with inclusion and have a workplace where those differences can thrive and be leveraged to empower the lives of our colleagues, patients and customers. We respect and celebrate the diversity of our people, their backgrounds and experiences and provide equal opportunity for all.