AI guides accurate design of patient-tailored combinatorial strategies for AML treatment

FIMM researchers together with clinical collaborators from Helsinki University Hospital have implemented an effective machine learning approach that identifies patient-customized anticancer drug combinations. The platform combines single-cell RNA-sequencing and drug sensitivity profiles, validated in patient-derived leukemic cells.

Cancer cell populations harbor a number of genetically and epigenetically heterogeneous subpopulations, each of which may be associated with a distinct cellular function or phenotype such as drug sensitivity or resistance. An ideal treatment regimen would target different cancer subpopulations present at the time of treatment, thus avoiding resistance, but not healthy cells, that would lead to toxic side effects.

To address this heterogeneity when designing combinatorial treatment regimens for cancer patients, researchers from the Institute for Molecular Medicine Finland FIMM, University of Helsinki, implemented a computational-experimental platform to rationally combine drugs that selectively inhibit multiple cancer-related dysfunctions or resistance mechanisms in individual patients.

The platform is based on combining information from two different information sources: single-cell RNA sequencing and drug sensitivity testing on patient-derived cells. The single-cell sequencing data provides high-resolution information about the different cell subpopulations present in the sample.

The first case study utilizing this approach was recently published in Science Advances. In this study, researchers from FIMM, BRIC and HUS joined their forces to demonstrate how the platform enables prediction of synergistic drug combinations for patients with acute myeloid leukemia (AML). The four patient samples tested each presented with different molecular backgrounds and drug resistance patterns.

Importantly, in subsequent laboratory experiments, the research team was able to demonstrate that many of the predicted drug combinations were shown to selectively inhibit AML cells and to avoid co-inhibition of normal cells, thereby reducing toxic effects and increasing the likelihood for clinical success.

"To the best of our knowledge, this is the first approach to systematically tailor personalized combinatorial regimens that takes into account both the molecular heterogeneity of cancer cells and the possible toxic effects of the drug combinations on healthy cells, with the aim to improve both combination efficacy and tolerability”, says FIMM PhD student Aleksandr Ianevski, the first author of the study.

The approach solves the critical experimental and computational challenges in the design of combinatorial regimens by effectively prioritizing, among the massive number of anticancer drug combinations, those that warrant further testing in scarce patient-derived primary cells. Since the platform uses only a limited number of patient primary cells, it is widely applicable to hematological and other cancers that are accessible for single-cell RNA-sequencing and drug sensitivity testing for novel precision oncology applications. 

“Our approach accurately predicted patient-specific combinations that resulted not only in synergistic cancer cell co-inhibition but that were also capable of targeting specific AML cell subpopulations that emerge in differing stages of disease pathogenesis or treatment regimens with close to real-time clinical timeframe”, says Dr. Anil K. Giri, co-corresponding author of the study.

The study originated from a collaboration between several research groups at FIMM, clinicians at the Helsinki University Hospital Comprehensive Cancer Center, under the Functional precision medicine in cancer grand challenge programme. The study made use of several FIMM core units including High Throughput Biomedicine Unit, Single-cell Analytics Unit, and Sample Storage Infrastructure.

Original publication: Aleksandr Ianevski, Jenni Lahtela, Komal K. Javarappa, Philipp Sergeev, Bishwa R. Ghimire, Prson Gautam, Markus Vähä-Koskela, Laura Turunen, Nora Linnavirta, Heikki Kuusanmäki, Mika Kontro, Kimmo Porkka, Caroline A. Heckman, Pirkko Mattila, Krister Wennerberg, Anil K. Giri and Tero Aittokallio. Patient-tailored design for selective co-inhibition of leukemic cell subpopulations, Science Advances Vol. 7, no. 8, eabe4038, DOI: 10.1126/sciadv.abe4038


Further information:

Anil K. Giri, PhD, postdoctoral researcher

Institute for Molecular Medicine Finland (FIMM), HiLIFE, University of Helsinki



Krister Wennerberg, PhD, professor

Biotech Research & Innovation Centre (BRIC), University of Copenhagen



Tero Aittokallio, PhD, professor

Institute for Molecular Medicine Finland (FIMM), HiLIFE, University of Helsinki