Research data management (RDM)

We are committed to the success of FAIR (Findable, Accessible, Interoperable, and Reusable) and open research data. Research data possess immense value, which, when combined with tomorrow's technologies such as machine and deep learning, can unlock answers to questions we cannot even conceive today. Therefore, the flexible contextualization of research data with machine-actionable metadata is essential to advance modern science.

We draw inspiration from the open-source software community's success, developing approaches that enable the biological community to collaboratively build a community-wide FAIR research data resource. Our research group participates in the National Research Data Infrastructure (NFDI) initiative with the DataPLANT project, which empowers plant researchers to engage in a thriving RDM ecosystem without barriers.

Biological data science

Modern technologies now allow researchers to simultaneously study biological organisms and processes across various molecular layers and in diverse biological contexts. To leverage this wealth of data, we are developing advanced methods for both supervised and unsupervised analysis of multi-modal omics data using cutting-edge machine learning and statistical modeling techniques. As part of fslabs.org, we contribute to the open-source data analysis library environment, ensuring that our innovations are accessible to the broader research community. Our methods enable comprehensive, data-driven integration and analysis of data derived from multiple omics technologies and varied biological scenarios, thereby enhancing our understanding of complex biological systems.

Prediction of acclimation modulators and integrators based on ‘omics data

Acclimation responses involve complex interactions among genes, proteins, and metabolites, resulting in both general and specific elements. Previously, we focused on extracting response-specific structural models from complex 'omics data using functional constraint aggregation and network topology inference.

Predicting modulators and integrators based on regulatory network topology and differential expression levels remains challenging, especially with limited acclimation data. However, incorporating additional regulatory characteristics of proteins—such as thermostability, aggregation propensity, intrinsic disorder, and post-translational modifications—can improve this identification.

This project proposes a novel approach using sequential meta-learning to integrate these regulatory characteristics, overcoming data sparsity challenges. In collaboration with CRC experimental groups, we will test our model’s predictions by introducing thermostable protein homologs into plant and algal cells. These experiments are expected to reduce acclimation capabilities and cause growth retardation, validating our approach in experimental evolution assays.