Water is essential for life, but some organisms can survive extreme dehydration, a phenomenon called anhydrobiosis. Orthodox plant seeds are remarkable examples, capable of enduring desiccation and surviving harsh conditions for years, even millennia. This ability allows seeds to remain dormant until favorable conditions arise. However, desiccation poses challenges, including oxidative stress and molecular damage, particularly to DNA. Damaged DNA can prevent germination and hinder development after rehydration, yet the mechanisms protecting DNA during these processes remain largely unknown. In this project, I aim to uncover the mechanisms that safeguard genome integrity in desiccated seeds and enable successful germination. By addressing this knowledge gap, we hope to improve seed vigor, longevity, and resilience in crops, contributing to agricultural productivity and global food security.
The FibForm project studies how plants can control the nanostructure of their cell walls, so that they obtain the desired strength and other properties. The idea behind my research is that the specific structure of hemicellulose polymers, a main component of plant cell walls, guides the building of cellulosic fibrillar structures. I plan to confirm this hypothesis by studying different kinds of plants and model systems, in which differences in the hemicelluloses result in different fibril morphologies. By better understanding the relationships between the chemical structure of the hemicelluloses and the resulting properties of the plant cell walls, we can develop plants for more sustainable agriculture.
Machine learning (ML) competitions aim to improve healthcare algorithms, like detecting lung cancer in chest x-rays. Teams worldwide compete to develop such algorithms, driven by prizes or prestige. Competitions encourage innovation but often produce similar algorithms that excel in one accuracy metric but fail with diverse real-world data.
Relying on one accuracy metric is insufficient for measuring algorithm quality, for example for rare patient cases. It leads to many similar algorithms with high environmental cost. Moreover, competition can deter women and minorities from entering or staying in data science.
I propose competitions which look at multiple metrics across different patient subgroups, with applications in chest and cardiovascular disease. Using the latest ML techniques we will develop methods to improve evaluation of algorithm robustness while reducing the carbon footprint. We will also study how competitions affect women and other groups in data science.
Imagine a world where generative AI (GenAI) exists free from any concerns for safety, security, privacy, trustworthiness, misinformation, or bias. We are far from this vision today, where negative consequences from GenAI, such as deepfakes, are ever present in the news cycle. A new threat to GenAI has emerged recently, as large language models can be attacked by malicious actors, leading to leakage of private data, manipulation of end-users, and even risks of medical misdiagnosis. For instance, an attacker can modify prompts to ‘trick’ a model into releasing private data. The project aims to draw upon the field of linguistics to mitigate such attacks, relying on the hypothesis that there are identifiable linguistic patterns in signals that attempt to negatively affect a GenAI model. If we can identify such patterns, we may have the key for safe and secure language-driven AI in the future.