(1024-A) Address infections by multi-drug resistant bacteria with artificial intelligence-empowered combination therapies
Monday, February 5, 2024
12:00 PM – 1:00 PM EST
Location: Exhibit Halls AB
Abstract: Antimicrobial resistance (AMR) has greatly undermined the efficacy of established therapeutics, primarily due to their overuse or misuse. Among the various resistant organisms, carbapenem-resistant Enterobacteriaceae (CRE) is of particular concern as carbapenems are currently the last-line antibiotics for its infections. There is an unmet need to develop novel drugs to overcome such widespread resistance since the process is too extended, usually dozens of years, to be readily available. As an alternative, FDA-approved drug combinations offer more accessible therapeutic options to shorten drug development pipeline and to circumvent resistant pathways. Clinical data has also provided compelling evidence that combination therapy is the preferred approach to monotherapy in managing CRE infections, especially in cases where patients are critically ill.
In this study, we utilised an artificial intelligence (AI)-based platform, IDentif.AI, to optimise combination therapies against 2 representative CRE isolates, E. coli C31 and K.pneumoniae ENT646. 12 drugs, intentional or repurposed, at 3 concentration levels were included to form an immense drug-dose searching space of 531,441 combinatorial therapy candidates. It is prohibitively large to interrogate with conventional therapy development approaches, such as the high throughput screening. Independent of any pre-existing data or mechanism-of-action, IDentif.AI was used to perform optimisation by testing only 155 out of 531,441 possibilities based on an orthogonal composite array design (OACD) and derive a correlational relationship between drug combinations and their respective efficacies (eg. %Inhibition), from which we obtained a ranked list of all combinations including the optimal ones.
The results demonstrated that IDentif.AI-pinpointed top combinations achieved maximal inhibition against both isolates. Notably, bleomycin/meropenem (BLM/MEM) was consistently the top ranked combination across both C31 and ENT646. To further interrogate the potential actionability of the combination, BLM/MEM was validated in 7 other CRE isolates from 3 species: K. pneumoniae ENT448, ENT1192, E. coli C165, C242 and E. hormaechei C36, C52, C254. This combination demonstrated broad-spectrum activity against all aforementioned isolates with promising inhibitory and bactericidal effects. More importantly, additional synergy analyses revealed that BLM and MEM synergistically interacted at different dose ratios leading to enhanced efficacies. Aside from BLM/MEM, IDentif.AI pointed to other effective combination therapies including levofloxacin in combination with sitafloxacin, which resulted in 100.00±0.25 %Inhibition against ENT646.
This study presented an alternative approach to accelerate the development of AMR-specific treatment strategies and potentially enrich our arsenal of treatment options in the face of surging antimicrobial resistance. IDentif.AI-pinpointed BLM/MEM combination demonstrated promising results in vitro against multiple isolates of CRE species. However, further validations including in vivo and preclinical studies may be helpful in clinical translation. Nonetheless, the platform and the combinations may contribute to future pandemic readiness and the accessibility of treatment options for the clinical communities.