Periodic Reporting for period 4 - DECIPHER (A computational framework to interpret the chemical language of the microbiome)
Période du rapport: 2025-06-01 au 2025-11-30
Humans, animals and plants are covered in microbes. Such microbiomes have a major impact on the health of their hosts and have been linked to traits like growth promotion, stress resilience, and diseases. However, the mechanisms underlying microbiome-host interactions remain poorly understood. Recent studies have shown that microbiome-associated phenotypes are often mediated by specific molecules, a ‘chemical language’ that enables microbes to interact with each other and with the host. The biosynthesis of these molecules is encoded in metabolic gene clusters (MGCs) that are subject to frequent horizontal transfer and are therefore highly strain-specific.
Current computational methods for analysing microbiomes largely focus on comparative taxonomic analyses and generic metabolism, and overlook this complex "chemical dialog". Indeed, no adequate methods are available to systematically study the roles of MGCs in microbiomes. At the same time, metabolomics data from microbiomes are full of ‘dark matter’: unknown molecules that cannot be traced to their producers. Here, I propose to develop the first comprehensive computational framework to study the chemical language of the microbiome.
DECIPHER aims to develop new algorithms to link MGCs to their metabolic products and to predict their molecular and ecological functions in microbiomes. Subsequently, the new framework will be applied in systematic studies of microbiome- associated phenotypes in plants and humans. Together, the innovations proposed here will fill a key gap in the analysis of microbiome function and pave the way toward precision-engineering of microbiomes with specific metabolic capabilities for designer soils and microbiome-based therapies.
During its runtime, DECIPHER succesfully developed a range of new computaitonal tools for the analysis of functions of biosynthetic gene clusters and their metabolic products in microbiomes. Moreover, they were applied in a prototypical experimental setting, where culture-dependent and -independent approaches were successfully integrated to identify genes and BGCs responsible for a complex phenotype, disease suppression in rhizosphere microbiomes.
Current computational methods for analysing microbiomes largely focus on comparative taxonomic analyses and generic metabolism, and overlook this complex "chemical dialog". Indeed, no adequate methods are available to systematically study the roles of MGCs in microbiomes. At the same time, metabolomics data from microbiomes are full of ‘dark matter’: unknown molecules that cannot be traced to their producers. Here, I propose to develop the first comprehensive computational framework to study the chemical language of the microbiome.
DECIPHER aims to develop new algorithms to link MGCs to their metabolic products and to predict their molecular and ecological functions in microbiomes. Subsequently, the new framework will be applied in systematic studies of microbiome- associated phenotypes in plants and humans. Together, the innovations proposed here will fill a key gap in the analysis of microbiome function and pave the way toward precision-engineering of microbiomes with specific metabolic capabilities for designer soils and microbiome-based therapies.
During its runtime, DECIPHER succesfully developed a range of new computaitonal tools for the analysis of functions of biosynthetic gene clusters and their metabolic products in microbiomes. Moreover, they were applied in a prototypical experimental setting, where culture-dependent and -independent approaches were successfully integrated to identify genes and BGCs responsible for a complex phenotype, disease suppression in rhizosphere microbiomes.
The project has made significant strides in understanding how microorganisms communicate and produce valuable compounds that hold immense potential for applications in medicine, agriculture, and beyond.
One of the key achievements of the project was the development of (prototypes of) innovative software packages, which leverage cutting-edge artificial intelligence techniques to predict the functions of biosynthetic gene clusters (BGCs). These packages have already unveiled fascinating insights into the roles of microorganisms in producing diverse chemical compounds, and have led to several collaborations with experimental scientists to confirm the predicted functions of these BGCs.
Moreover, the creation of an in silico retrobiosynthesis software may enables us and other researchers to identify the building blocks encoded within BGCs, and identify structure-function relationships that connect specific chemical substructures to functions. New collaborations with industry partners have been instrumental in collecting a vast library of almost 4000 purified natural products. This treasure trove of compounds serves as a valuable resource for cell painting assays, allowing us to predict how these compounds may interact with cells.
The isolation of hundreds of strains from disease-suppressive soils and subsequent genome sequencing further helped us to validate our approaches. We performed paired metagenomics/metatranscriptomics/metabolomics analyses of the synthetic communities associated with the disease-suppressive phenotype, and identified multiple genes associated with fungal inhibition, with one BGC product being validated with this function in vitro.
One of the key achievements of the project was the development of (prototypes of) innovative software packages, which leverage cutting-edge artificial intelligence techniques to predict the functions of biosynthetic gene clusters (BGCs). These packages have already unveiled fascinating insights into the roles of microorganisms in producing diverse chemical compounds, and have led to several collaborations with experimental scientists to confirm the predicted functions of these BGCs.
Moreover, the creation of an in silico retrobiosynthesis software may enables us and other researchers to identify the building blocks encoded within BGCs, and identify structure-function relationships that connect specific chemical substructures to functions. New collaborations with industry partners have been instrumental in collecting a vast library of almost 4000 purified natural products. This treasure trove of compounds serves as a valuable resource for cell painting assays, allowing us to predict how these compounds may interact with cells.
The isolation of hundreds of strains from disease-suppressive soils and subsequent genome sequencing further helped us to validate our approaches. We performed paired metagenomics/metatranscriptomics/metabolomics analyses of the synthetic communities associated with the disease-suppressive phenotype, and identified multiple genes associated with fungal inhibition, with one BGC product being validated with this function in vitro.
The DECIPHER project unlocked new horizons in microbiome science and natural product research, generating valuable insights into the chemical language of the microbiome.
During the last phase of the project, our software packages were further refined. Through iterative learning and algorithmic improvements, we achieved enhanced accuracy and efficiency in predicting BGC functions and identifying substructures within BGCs data. These developments deepened our understanding of the complex interactions between microorganisms and their chemical products and opened new opportunities for drug discovery and biotechnological applications.
The collaboration with industry partners yielded important outcomes through the application of cell painting assays to predict the mechanisms of action of specialized metabolites. The extensive dataset generated during the project, combined with machine learning approaches, enabled the discovery of structure–function relationships and provided insight into the therapeutic potential of these compounds. These efforts demonstrated clear potential to reshape drug development strategies and inform future pharmaceutical research.
A major focus of the project was the functional dissection of disease-suppressive rhizosphere microbiomes. Although certain bacterial strains proved difficult to isolate, the cultivated community — representing a substantial portion of the natural microbiome — provided a strong basis for investigation. Through detailed analysis of isolated strains and their phenotypic characteristics, and by using synthetic microbial communities together with metatranscriptomic and metabolomic analyses, we uncovered mechanisms underlying disease suppression and identified promising leads for sustainable biocontrol strategies in agriculture.
Overall, the project delivered substantial breakthroughs in our understanding of the microbiome’s chemical language and the hidden potential of microorganisms. The integration of multidisciplinary approaches drove innovation and resulted in impactful scientific and application-oriented outcomes.
During the last phase of the project, our software packages were further refined. Through iterative learning and algorithmic improvements, we achieved enhanced accuracy and efficiency in predicting BGC functions and identifying substructures within BGCs data. These developments deepened our understanding of the complex interactions between microorganisms and their chemical products and opened new opportunities for drug discovery and biotechnological applications.
The collaboration with industry partners yielded important outcomes through the application of cell painting assays to predict the mechanisms of action of specialized metabolites. The extensive dataset generated during the project, combined with machine learning approaches, enabled the discovery of structure–function relationships and provided insight into the therapeutic potential of these compounds. These efforts demonstrated clear potential to reshape drug development strategies and inform future pharmaceutical research.
A major focus of the project was the functional dissection of disease-suppressive rhizosphere microbiomes. Although certain bacterial strains proved difficult to isolate, the cultivated community — representing a substantial portion of the natural microbiome — provided a strong basis for investigation. Through detailed analysis of isolated strains and their phenotypic characteristics, and by using synthetic microbial communities together with metatranscriptomic and metabolomic analyses, we uncovered mechanisms underlying disease suppression and identified promising leads for sustainable biocontrol strategies in agriculture.
Overall, the project delivered substantial breakthroughs in our understanding of the microbiome’s chemical language and the hidden potential of microorganisms. The integration of multidisciplinary approaches drove innovation and resulted in impactful scientific and application-oriented outcomes.