DynamicKit

Finding new combination therapies against multi-resistant tuberculosis through a new proteomics technology and artificial intelligence
Artificial intelligence-assisted translation of a new bioassay to decipher the dynamic mode of action of tuberculosis-active antibiotics to develop new combination therapies for multidrug-resistant and dormant tuberculosis.

Tuberculosis (TB) is the deadliest infectious disease in humans and claims around 1.5 million lives worldwide every year. To successfully treat this lung disease, a mix of different drugs has to be administered for several months. This however, is problematic as the bacterial pathogens become resistant, and even in treatable bacterial populations, highly resistant subpopulations can be detected. In order to prevent the spread of the disease, it is therefore not only necessary to develop new antibiotics, but also to constantly find new combinations of active substances. Such combinations can so far only be identified empirically in expensive clinical studies. New digital tools in combination with novel analytical tools, such as artificial intelligence self-learning algorithms, have the potential to decipher the interplay of different antibiotics on mycobacterial metabolism in a faster, more cost efficient way, making it possible to identify suitable drug cocktails to overcome TB drug-resistance and improve current treatment regimens.

Multi-drug-resistance (MDR) is a dramatic challenge for the most deadly infectious disease of the world, tuberculosis (TB). In our project, we use self-learning algorithms to understand the interaction of different drugs in their effect on the metabolism of mycobacteria, the causative agents of tuberculosis. This way, we are not only able to predict new suitable drug combinations for tuberculosis treatment, but also determine biological molecules that reflect resistance mechanisms, so that we can find out how we can specifically reverse this with drugs. This combined approach yields an urgently needed preclinical laboratory model that will enable us to stop the further spread of the disease.

Strategy and conditions

The identification of new drug-combinations in TB is very difficult as appropriate preclinical models to predict synergistic effects are missing, so that is the unmet need we plan on addressing.

First, we will study the action of common antimycobacterials. A new experimental technique developed at the LMU allows us to describe their modes of action, escape mechanisms and adaptive reactions over time in unprecedented detail. This will enable us to characterize the effect of different antibiotics both individually and in combination.

This analysis will be extended thanks to artificial intelligence and systems biology. The obtained dynamic data will be modelled to gain a better understanding of the pathogen and ultimately highlight novel drug targets. Neural-networks and random forests can be used to perform in silico screens of untested drug combinations, predicting their impact.

Aims of the research project

Our main goal is to develop fundamentally new approaches against resistant as well as susceptible tuberculosis leveraging the potential of new experimental methods and artificial intelligence.

In this way, we hope to find out which active ingredients are an ideal match to be used as combination treatment for tuberculosis. This could provide less toxic and shorter treatment regimens. Furthermore, we aim to identify new drug combinations that may efficiently battle drug-resistant TB.

Expected benefits for society

According to the WHO, antimicrobial resistances such as the ones we find in tuberculosis currently pose the greatest long-term threat to human health and wellbeing. This project builds on novel, data science approaches within basic research to address and counteract the development and spread of resistance within this infectious disease.

Team

PD Dr. Andreas Wieser research group is spearheading a new proteomic technology, which for the first time enables to accurately measure newly formed proteins over time in Mycobacteria, the causative agents of tuberculosis. Prof. Dr. Michael Hoelscher contributes with his world leading expertise in infectious diseases and experience in coordinating drug trials. Through his work group we have access to novel substances and data on clinical correlates of diseases and treatment. Prof. Dr. Dr. Fabian Theis and Dr. Michael Menden are driving computational analyses with artificial intelligence.

PD Dr. Andreas Wieser
Project Management

Ludwig-Maximilians-Universität München
Max von Pettenkofer Institut

Prof. Dr. Fabian Theis
Project Management

Technische Universität München
Institut für Computational Biology
Helmholtz Zentrum München

Prof. Dr. med. Michael Hoelscher
Project Management

Ludwig-Maximilians-Universität München
Department of Infectious Diseases and and Tropical Medicine
Medical Faculty

Dr. Michael Menden
Project Management

Ludwig-Maximilians-Universität München
Institut für Computational Biology
Helmholtz Zentrum München

Ludwig-Maximilians-Universität München
Max von Pettenkofer Institut

Technische Universität München
Helmholtz Zentrum München

Ludwig-Maximilians-Universität München
Department of Infectious Diseases and and Tropical Medicine
Medical Faculty

Technische Universität München
Institut für Computational Biology

Cooperations

With their interdisciplinary basic research, the research team combines expertise from fields such as bioinformatics, artificial intelligence and machine learning to understand cellular processes (F. Theis / M. Menden), analytical chemistry, medical microbiology (A. Wieser) and tropical medicine, including therapy in tuberculosis/clinical trials (M.Hoelscher). This project will strongly benefit from the scientific network provided by BayResQ.net as well.

Publications

Identifying dormant growth state of mycobacteria by orthogonal analytical approaches on a single cell and ensemble basis.
Neumann AC, Bauer D, Hölscher M, Haisch C, Wieser A.
Anal Chem. 2018 Dec 3. doi: 10.1021/acs.analchem.8b03646. IMPACT 6,3 (2016)

High-dose rifampicin, moxifloxacin, and SQ109 for treating tuberculosis: a multi-arm, multi-stage randomised controlled trial.
Boeree MJ, Heinrich N, Aarnoutse R, Diacon AH, Dawson R, Rehal S, Kibiki GS, Churchyard G, Sanne I, Ntinginya NE, Minja LT, Hunt RD, Charalambous S, Hanekom M, Semvua HH, Mpagama SG, Manyama C, Mtafya B, Reither K, Wallis RS, Venter A, Narunsky K, Mekota A, Henne S, Colbers A, van Balen GP, Gillespie SH, Phillips PPJ, Hoelscher M; PanACEA consortium.
Lancet Infect Dis. 2017 Jan;17(1):39-49. doi: 10.1016/S1473-3099(16)30274-2. Epub 2016 Oct 26.
Read the article

Assessment of the sensitivity and specificity of Xpert MTB/RIF assay as an early sputum biomarker of response to tuberculosis treatment.
Friedrich SO, Rachow A, Saathoff E, Singh K, Mangu CD, Dawson R, Phillips PP, Venter A, Bateson A, Boehme CC, Heinrich N, Hunt RD, Boeree MJ, Zumla A, McHugh TD, Gillespie SH, Diacon AH, Hoelscher M; Pan African Consortium for the Evaluation of Anti-tuberculosis Antibiotics (PanACEA).
Lancet Respir Med. 2013 Aug;1(6):462-70. doi: 10.1016/S2213-2600(13)70119-X. Epub 2013 Jul
Read the article

Community assessment to advance computational prediction of cancer drug combinations in a pharmacogenomic screen.
Menden MP, Wang D, Mason MJ, Szalai B, Bulusu KC, Guan Y, Yu T, Kang J, Jeon M, Wolfinger R, Nguyen T, Zaslavskiy M; AstraZeneca-Sanger Drug Combination DREAM Consortium, Jang IS, Ghazoui Z, Ahsen ME, Vogel R, Neto EC, Norman T, Tang EKY, Garnett MJ, Veroli GYD, Fawell S, Stolovitzky G, Guinney J, Dry JR, Saez-Rodriguez J.
Nat Commun. 2019 Jun 17;10(1):2674. doi: 10.1038/s41467-019-09799-2. PubMed PMID: 31209238; PubMed Central PMCID: PMC6572829.

Deep learning: new computational modelling techniques for genomics.
Eraslan G, Avsec Ž, Gagneur J, Theis FJ.
Nat Rev Genet. 2019 Jul;20(7):389-403. doi: 10.1038/s41576-019-0122-6. Review. PubMed PMID: 30971806.