Integrative systems biology

Proteins mediate cellular functions by interacting with each other and with other biomolecules such as DNA, RNA, or metabolites. Thus, protein interactions determine the function of a gene and its protein product. The protein interactome describes the entirety of all protein interactions in a cell, and underlies genotype-to-phenotype relationships. Deciphering the cellular contexts under which protein interactions take place, how these interactions form, and what their functional outcome is remains inherently difficult. Yet, this knowledge is required to accurately predict how genetic mutations, toxins, and pathogens alter protein interactions and thereby perturb cellular function and cause disease (Figure 1).

Understanding genotype-to-phenotype relationships starts with identifying the underlying interactome. However, while the genomes, transcriptomes, and proteomes of entire human populations down to single cells have now been mapped, the mapping of protein interactomes remains very challenging. More recently, various major research efforts have produced the first systematic, proteome-wide maps of the human protein interactome (Figure 2, Luck et al. Nature 2020, Huttlin et al. BioRxiv 2020, Luck et al. TiBS 2017). However, due to technical limitations, these resources do not provide information on the molecular function, mode of binding, or cellular context of the interactions (Figure 2).

The aim of my lab is to develop integrative computational and experimental strategies that efficiently annotate protein interactions with information on their structure, function, and cellular context. We will then apply these approaches to predict and validate the molecular mechanisms of genome maintenance processes that are perturbed in neurodevelopmental disorders.

We are a highly interdisciplinary and collaborative lab, integrating omics data analyses with structural and network biology to build predictive models of molecular mechanisms, which we then validate using biophysical and cellular approaches. We favour data-driven approaches to make novel biological discoveries.


Prediction and experimental validation of interfaces in protein interactions

Identifying the regions and specific residues that mediate an interaction between two proteins is critical to predict which mutations will perturb the interaction, which interaction partners have overlapping binding regions (and are therefore mutually exclusive), and which interactions can co-exist (and thus mediate the assembly of larger protein complexes).

Furthermore, identifying binding regions often provides additional information on the molecular function of the interaction. Experimental characterisation of protein interaction interfaces often involves mutational scanning or protein fragmentation, which are both laborious and highly error-prone due to possible nonspecific perturbation of the entire protein fold. Many protein interactions are mediated by globular domains that either bind other types of globular domains or short stretches of amino acids (so-called linear motifs) that occur in disordered regions of proteins (Figure 3). We develop novel approaches to predict known types of domain-domain and domain-linear motif interfaces, discover novel types of interfaces in known protein interactions, and seek to experimentally validate these predictions by designing and testing compensatory mutations in the predicted interaction interface.

Prediction and experimental validation of the effect of mutations on protein interactions

By comparing the genomes or exomes of healthy and diseased individuals, scientists can sometimes statistically associate genetic variants with phenotypes. However, because many Mendelian-like diseases are caused by extremely rare mutations, most of the genetic variants identified in these studies remain uncharacterised. These so-called variants of uncertain significance are exponentially accumulating in variant databases (Figure 4), hindering genetic diagnosis, identification of disease mechanisms, and development of therapies. Contrary to common belief, mutations in proteins often only partially perturb their function, i.e. by disrupting some interactions while leaving others intact (Sahni et al Cell 2015). We develop computational and experimental approaches to predict and validate the effect of mutations on protein interactions and predict the phenotypic effect of uncharacterized variants by comparing interaction profiles of pathogenic and benign variants (Figure 5, Yadav et al. Curr Opin Biotech 2020).


Development of CL-MS for medium-throughput mapping of protein interaction interfaces

Cross-linking mass spectrometry (CL-MS) has emerged as a powerful technique to map protein interactomes, decipher protein complex topology, and finely-map interfaces in protein interactions. However, it remains unclear which combinations of cross-linking reagents can be used to map different interaction interfaces, whether some interaction interfaces are more or less amenable to detection with CL-MS, and whether protocols can be generalised to increase mapping throughput. We build protein interaction reference datasets and use them to develop and benchmark CL-MS protocols for systematically characterising protein interaction interfaces. We are collaborating with CL-MS experts and the in house proteomics Core Facility to achieve this goal.

Characterising the mutational landscape of proteins implicated in neurodevelopmental disorders

Neurodevelopmental disorders (NDDs) comprise various syndromes such as intellectual disability, developmental delay, autistic-like behaviour, and epilepsy. Often, NDDs are caused by de novo mutations in the germline, making them extremely hard to statistically associate with disease. Thus, despite sequencing thousands of exomes, the majority of children suffering from NDDs remain without a genetic diagnosis and it is difficult to understand the molecular causes of NDDs. Yet, ever-growing resources of human genetic variation also provide unique opportunities to study NDDs. We seek to apply our computational and experimental approaches to structurally and functionally characterise the mutational landscape of proteins implicated in NDDs in collaboration with local and international neurobiologists and clinicians. 

Understanding the brain specificity of neurodevelopmental disorders

Interestingly, of the many genes that have so far been linked to NDDs when mutated, more than 90% do not have a brain-specific expression pattern. This suggests that the brain-specific phenotype caused by mutations in these proteins must result from perturbations within their brain-specific network context. We are integrating brain transcriptome and proteome data with protein interaction resources to predict the molecular mechanisms responsible for the brain-specific phenotypes of NDD-associated proteins that do not have brain-specific expression patterns. We will then test whether pathogenic but not benign mutations interfere with the predicted mechanisms in vitro and in vivo. We will focus in particular on proteins that are implicated in genome maintenance, in order to elucidate novel mechanisms that adapt genome maintenance processes to brain-specific contexts.


I am a fellow of the DFG’s Emmy Noether Program for young investigators, who have granted me €1.6m in funding to support my research ambitions for the coming 6 years. These funds are topped up with IMB core funding.


I am currently accepting applications for:

  • a PhD position in the field of integrative network biology and interaction profiling (please apply via the IMB PhD programme)
  • internships of at least 2 months (i.e. as part of the International Summer School programme of IMB)
  • Bachelor and Master students (please apply via email)