Maximillien FIL's thesis “Design and optimisation of new natural drugs to combat tuberculosis”.”

Mr Maximillien FIL , doctoral student at IMBE, team Functional ecology: from socio-ecological systems to molecules, will publicly defend his thesis on Thursday 27 November 2025 at 2pm in the Thesis Room of the Faculty of Pharmacy.

In front of a jury made up of :

• Professor Nicolas WILLAND, Rapporteur, University of Lille
• Professor Rino RAGNO, Rapporteur, Sapienza University of Rome
• Doctor Jessica HERNANDEZ, Examiner, Aixial
• Professor Laurence FERAY, Chair of the Jury, Aix-Marseille University
• Doctor Sandrine ALIBERT, Doctoral supervisor, Aix-Marseille University
• Professor Gérard BOYER, Co-supervisor, Aix-Marseille University
• Doctor Paul BREMOND, Guest Member, Aix-Marseille University
• Professor Christine CONTINO-PEPIN, Guest Member, University of Avignon

Summary of work:

Since ancient times, tuberculosis has been one of the main causes of death affecting mankind. Since the second half of the 20th century^e Although the disease is increasingly being treated, it is still the leading cause of death from infectious diseases worldwide. In addition, the emergence of resistance to the pathogen Mycobacterium tuberculosis, In the face of these challenges, the development of new therapies is essential, particularly for rifampicin and isoniazid - two of the four molecules in the quadritherapy used as first-line treatment.

One of the first obstacles to the development of new anti-tuberculosis therapies is the mycobacterial membrane, a particularly complex and lipophilic vital membrane that represents a real barrier to antibiotics. Among other things, this membrane is characterised by the presence of mycolic acids and arabinogalactan. Destabilising this membrane is now a method of choice in the development of new anti-tuberculosis compounds.

In order to contribute to the fight against the disease, this study proposes the development of 3D-QSAR (Quantitative 3D Structure-Activity Relationship Analysis) models based on 3D pharmacophoric models as well as models of machine learning, and in particular Random Forest (RF) for the discovery of compounds capable of targeting Enoyl-[acyl-carrier-protein] reductase (InhA) and Decaprenylphosphoryl-beta-D-ribose oxidase (DprE1), two enzymes involved in the biosynthesis of mycolic acids and the mycomembrane of Mycobacterium tuberculosis. Forty-three 3D pharmacophoric models were produced, five of which were selected for the development of RF. 10 models of RF were then carried out, with F1 scores ranging from 0.88 to 0.97 and accuracy scores between 0.96 and 0.99.

Analysis of these models highlights the importance of descriptors such as the Shape Index, molar mass and globularity in identifying new InhA inhibitors. Similarly, while Shape Index and molar mass remain key for DprE1 inhibition, other parameters such as Van der Waals volume and surface area or LogP must be taken into account to inhibit this target correctly.

These models were then used for screening in silico of nearly 30,000 compounds, mostly of natural origin. During the initial screening in vitro she first 16 compounds selected and tested on Mtb H₃₇Ra, 7 showed a MIC₅₀ i.e. a hit selection rate of 47 %.

Among the potentially active compounds, 7 compounds were predicted to be active by all the models for the two targets InhA and DprE1. Among these compounds was a quinazolin-4-one substituted in positions 3, 6 and 7. As a «multitarget» approach could be a key solution in preventing the emergence of resistance, it was decided to synthesise a series of 17 quinazolin-4-ones substituted in position 3 by a peptide chain whose microbiological tests on Mtb H₃₇Ra are in progress.

All in all, this work shows the potential of 3D-QSAR models and particularly of machine learning in the fight against tuberculosis and in the process of Drug Discovery. They also highlighted the anti-tuberculosis potential of quinazolin-4-ones substituted in position 3 by a peptide chain.