Student for improving predictive models with additional synthetic transcriptomes

Reageer t/m 12 mei

Join a translational AI-oncology project that uses atlas-scale bulk, single-cell and spatial transcriptomic data to generate synthetic transcriptomes as realistic as possible and improve predictive models of cancer response.

What will you do?

In this internship you will contribute to a computational proof-of-concept that improves predictive model performance by using additional synthetic transcriptomes.

Your activities may include:

• Accessing the already harmonised 800,000 bulk, over millions single-cell and >1,000 spatial transcriptomic datasets
• Collect and gather immunotherapy response datasets which have transcriptomes along with clinicopathological information.
• Obtain the performance of the models with the transcriptomes present currently.
• Use state of the art generative models to generate realistic synthetic transcriptomes for each context and measure the similarity of the transcriptomes.
• Thereafter add the synthetic transcriptomes to the predictive model training and retrain.
• Compare and characterize the performance with and without additional synthetic transcriptomes.
• Evaluating model outputs for robustness, biological plausibility and reproducibility across cohorts
• Visualising results and translating findings into a scientific report and/or manuscript
• Contributing to reproducible code, documentation and analysis pipelines

What is this research about?

You will work within UMCG on a multidisciplinary project at the intersection of medical oncology, computational immunology, spatial transcriptomics and AI. The project builds on unique harmonised datasets from endometrial cancer, including bulk, single-cell and spatial transcriptomic profiles, and is embedded in a network spanning oncology, pathology, molecular biology and biomedical AI.

Not all cancer patients respond to the immunotherapy treatments. Predictive models do not perform well due to low sample size in rare cancers. Can generating realistic synthetic transcriptomes help improve the performance of the predictive models?

Whom are we looking for?

You are a MSc or advanced BSc student in bioinformatics, computational biology, AI, data science, biomedical sciences, mathematics or a related field.

• Experience with Python and/or R for data analysis
• Affinity with machine learning, statistics or generative modelling
• Interest in cancer immunology, transcriptomics or spatial omics
• Ability to work with large and heterogeneous datasets
• Strong analytical thinking and problem-solving skills
• Clear written and verbal communication skills
• Independent, curious and motivated, while also enjoying collaboration in a multidisciplinary team

What we offer

• Internship agreement with UMCG
• Intensive supervision in a multidisciplinary research environment
• Opportunity to work on a novel AI and computational immunology project with atlas-scale data

Good to know: in consultation, part of the internship can be performed from home.