Research assistants wanted: Deep generative models for spatial reconstruction of single cell transcriptomics data
We are looking for two ambitious master students for two exciting, paid research assistant positions. The positions are for 1 day/week for 10 months, ideally to be combined with a master thesis research project. Project start: October
The spatial organization of cells underlies the function of many biological systems, including tumors. Emerging biotechnologies now allow for spatially resolved quantification of transcriptomes at single cell resolution , but are expensive and low-throughput. On the other hand, conventional single cell RNA-sequencing is now routinely applied to large sample cohorts, but does not retain spatial information. Recently, spectacular advances in generative models  has led to models such as ImaGen  and DALL-E , that can synthesize high-quality and high-fidelity images from text or metadata.
In this project, we will employ deep generative models to reconstruct the spatial structure of biological samples from single cell transcriptomic count data alone. To this end, we will formulate and train a deep generative model on spatial transcriptomics data, matched with single cell count data. The trained models will then be employed to synthesize “images” using count data only as input metadata, which will allow us to extract biologically meaningful spatial information (e.g. extent of immune infiltration in tumor) from the images.
This is a high-risk, high-gain project. If successful, it will allow for augmentation of single cell transcriptomics dataset with spatial features, adding great value to such datasets.
● Background in AI (programming in Python and PyTorch, the basics of mathematics and deep learning) or molecular biology.
● Strong communication skills
To apply, send an email to Jakub Tomczak (email@example.com)) and Evert Bosdriesz (firstname.lastname@example.org) with a brief motivation and grade list.
- Moses L, Pachter L. Museum of spatial transcriptomics. Nat Methods. 2022;19: 534–546.
- Tomczak JM. Deep Generative Modeling. 2022. doi:10.1007/978-3-030-93158-2
- Saharia C, Chan W, Saxena S, Li L, Whang J, Denton E, et al. Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding. arXiv [cs.CV]. 2022. Available: http://arxiv.org/abs/2205.11487
- Ramesh A, Dhariwal P, Nichol A, Chu C, Chen M. Hierarchical Text-Conditional Image Generation with CLIP Latents. arXiv [cs.CV]. 2022. Available: http://arxiv.org/abs/2204.06125
MSc project: Investigating how human and algorithmic agents develop novel modes of intelligence
One of the areas where life-and-death decisions are increasingly made based on algorithmic technologies (e.g., deep-learning) is “medical diagnosis”. Every day, doctors who are used to working with their eyes and their traditional technologies, are confronted with new algorithmic technologies (such as wearable devices, AI-empowered diagnosis tools, and many algorithms that offer them insights into medical issues). The situation is at best like giving an advanced car to those who are used to riding bikes! You can imagine novel expertise and experiences emerge when these professional cyclists find themselves behind the driving panel of these rather autonomous, unknowable AI technologies. What happens to their expertise? How effectively can they use their medical expertise? How and under which conditions might they fall into certain ineffective ways of working such as automation bias, confirmation bias, ruling-out bias against the technology, and many new traps that we just do not know?
In this research project, which is part of an ongoing research program at KIN center for digital innovation, we engage medical professionals in a range of experimental procedures, where they work with various forms of AI tools and under different working conditions. This way, we examine how they use and develop their expertise for medical diagnosis when working with these technologies. The results offer developers of these algorithms the chance to proactively learn about the (in)effective design choices and the (un)intended consequences of their technologies. It also helps professionals from medical and (partly) other domains of knowledge work learn how they can develop novel expertise of working with these technologies. Finally, we gain insights about how to assess, organize, and regulate these technologies in order to minimize their ineffective modes of operation.
Are you interested in developing your research (thesis) in this domain? This would be an opportunity for those who are interested in advancing our knowledge of human-technology intelligence and expertise, by combining both technical and social aspects. If you are interested in knowing more about this research project, please send an email to Mohammad H. Rezazade Mehrizi (email@example.com), School of Business & Economics, Knowledge, Information and Networks Research Group.
Deadline: 15th January, 2022
Do you want to contribute to the next generation of healthcare fueled by Artificial Intelligence? We then invite you to apply for the position of Associate professor for AI and Health, a joint position of the Computer Science Department of the VU and the Cancer Center Amsterdam of the Amsterdam UMC. Please find more information here.
Deadline: 17th January, 2022
We are looking for a new PhD student who will do research in the area of Machine Learning (ML) and in particular on supervised learning and Reinforcement Learning (RL). The position is embedded in the Quantitative Data Analytics group.
We are interested in attracting a PhD student able to perform ground-breaking research in fundamental aspects of Machine Learning and Reinforcement Learning. The PhD student will apply these novel algorithms in the context of an EU funded project called ICARE4OLD focused on improving care for elderly by machine learning based recommendations on interventions. For more information and to apply, please see here.