Ongoing projects
Intelligent drug identification for pandemic viruses
Until now, there is no established therapy against SARS-CoV-2. RNA viruses acquire mutations that can lead to resistance to direct-acting antiviral drugs such as remdesivir. In turn, viruses need key processes of the host cell comprising endocytosis, signalling, transcription and cytoskeletal remodeling. Antivirals targeting host proteins required for the viral life cycle are less likely to result in the selection of resistant viruses. Hence, a promising alternative can be to target host factors, which are essential for the virus’ life cycle. Repurposing clinically approved drugs or develop treatments with drug combinations advantages from the large compendium of approved drugs tipping a huge variety of disease causing proteins. Only recently, several Crispr/Cas9 screenings have been presented. However, there is only very limited consensus among these previous screening results. In this project, we aim to identify host dependency factors being essential for SARS-CoV-2 entry, replication and spread, and identify drugs and drug combinations targeting these host factors for effective prevention and treatment. We use data from an up-scalable screening method employing High Density Cell Arrays developed in the laboratory of Dr. Erfle for gene perturbation screens. Coronavirus infection experiments are conducted in the laboratory of Dr.med Tuna Toptan Grabmair at the Institute of Medical Virology, Frankfurt (Director: Prof. Dr. Sandra Ciesek). Our part is to identify host dependency factors employing machine learning, similarly, as we did in a pilot study to identify host dependency factors of bacteria or protozoans in D. melanogaster [1], to associate the phenotypes of protective knockdowns and to group the effects into cellular processes/pathways.
iRECORDS (International - Rapid rEcognition of CORticosteroiDs sensitivity or resistance in Sepsis)
Sepsis and COVID-19 are both placing a major burden on societies and populations worldwide. Deregulated host response to infection is the hallmark supporting the routine use of corticosteroids (CS), a low-cost and highly efficient class of immuno-modulators, in sepsis/COVID-19. Stratifying patients based on their individual immune response may improve the balance of benefit to risk of CS treatment. Only recently, we identified the ratio of IFNgamma/IL10 as a good biomarker for this [2]. This project aims to integrate data of DNA, RNA, proteins such as cytokines and hormones, or metabolite compounds to define the CS sensitivity/resistance of individual patients. Partners of all over Europe are involved. On our part we contribute to methods of artificial intelligence integrating the high dimensional multi-level data exploring gene networks.
Identifying and validating insecticidal targets against Plasmodium falciparum infected Anopheles gambiae
In 2019, appr. 229 million cases of malaria occurred worldwide, of which the majority were in the sub-Saharan region. This disease is caused by the parasite Plasmodium falciparum, transmitted by the mosquito Anopheles gambiae. In patients and in the vector, plasmodium must complete a complex developmental cycle before being passed from the patient to a newly infected individual. Malaria is treated by small molecule drugs affecting diverse developmental steps of the parasite, but with increasing resistance. Instead of treating the infection directly, an alternative is to control the mosquito. However, particularly children are susceptible to neurotoxicity of insecticides, so there is a need for new insecticides being not toxic to human.
Besides this, machine learning enables to incorporate heterogeneous data and training on complex data structures. However, to employ such machines successfully, a solid cell biological and molecular data background is mandatory. Essential genes are, by definition, indispensable for the life cycle of an organism and blocking them by e.g. specific drugs harms the population or cells leading to their eradication. Experimental screens for identifying essential genes by knockout or knockdown assays have been widely performed for a broad range of model organisms and assembled in databases. Disappointingly, when comparing the results, they show only little overlap. We are approaching this problem by data integration across organisms employing a large compendium of descriptors/features and machine learning covering the data from the main model organisms and their genomes, transcriptomes, homology/gene conservation and cellular networks [3,4]. This joint project is a collaboration with Prof Adebiyi from Covenant University, Nigeria, in which we are contributing to the bioinformatics and machine learning analysis, and Prof. Adebiyi’s lab is contributing in bioinformatics and the experimental validation.
Developing an individualised treatment approach for antibiotics therapy for patients hospitalised with moderate community-acquired pneumonia (CAP)
The role of macrolides, β-lactams and fluoroquinolones in CAP is controversially discussed. We aim to identify patient characteristics with which patients can be identified benefitting from a specific antibiotic therapy to implement this later into the clinical routine. We apply machine learning/artificial intelligence concepts to data available on admission of hospitalised patients from the observational, prospective, multinational CAPNETZ study [5] to investigate patient variables of all relevant aspects for CAP treatment and management based on appr. 12,000 patients. In a pilot study, we identified a treatment rule for macrolide/beta-lactam combination therapy in respect to a beta-lactam mono therapy [6].
Testing the microbiological effectiveness of UV-C radiation to break the infection chain in neonatal incubators
In Germany, each year 63,000 children are born too early. Worldwide, nearly one out of ten children are given prenatal birth (incidence: 9.7%). By this, premature infants are the largest group of patients in the peadiatrics. The improvements in cure and treatment options of prenatal infants have considerably contributed to a higher survival rate. Particularly very early born prenatal infants (Very Low Birth Weight, VLBW) are at high infection risk due to their immature organ system and immune response. Besides this, there is a high selection pressure to pathogens at nosokomial conditions. Prenatal infants are placed in incubators in humid and warm conditions (32-34 0C, 60-80% humidity). This environment unfortunately serves also germs ideal growing conditions. Initial microbiological investigations at a neonatal intensive care unit (NICU) showed that Gram positive and Gram negative bacteria were detected particularly at objects close to the patients including incubators.
The aim of our project is to microbiologically evaluate the performance of a novel device which disinfects the surface of the interior of incubators based on radiation by UVC-LEDs. A laboratory model developed by collaboration partners (SAVUNA GmBH Augsburg, Fraunhofer Institute IOSB Ilmenau, Micro-Hybrid Hermsdorf) will be tested within a clinical study. Our part in this project is the data analysis, and particulalrly the analysis of the sequenced microbiomes.
