Mobile Menu

MEDUSA: AI Used in Battle Against Mesothelioma

International genomics research, led by the University of Leicester, has used an artificial intelligence model – MEDUSA – to study mesothelioma tumorigenesis and identify potential areas for therapeutic intervention.

Malignant Pleural Mesothelioma

Malignant Pleural Mesothelioma (MPM) is a rare, incurable cancer with a poor prognosis. The incidence of MPM is increasing despite its primary cause, asbestos exposure, being characterised over 50 years ago. MPM typically grows to enormous volumes, averaging 640ml, and exhibits large variations in tumour aggressiveness. Overall, MPM has seen only limited drug development over the last two decades, with the exception of combination immune checkpoint inhibition. Personalised treatments for MPM are lacking, but recent insights into inter-patient genomic heterogenicity has revealed frequent somatic alterations involving the cancer genes BAP1, NF2 and CDKN2A. In addition to that, preclinical models have implicated these tumour suppressors as initiators of MPM development, with prognostic significance, particularly when they are concurrent.

Despite this, major questions remain regarding the clinical significance of genomic intratumour heterogeneity (ITH) in MPM. More specifically, whether constrained, temporal ordering of driver events is required for MPM transformation; critical information that could help to inform the development of precision therapies.

Previous clinical phylogenetic studies, such as TRACERx lung, have shown that clonal neoantigen architecture drives immune surveillance, which is a selection pressure leading to immune evasion via immunoediting. However, it is not yet known the extent to which ITH modulates host immune surveillance or immune escape.

Mesothelioma evolution: deciphering drugable Somatic alterations (MEDUSA)

In this study, the researchers hoped to address the gaps in knowledge of ITH in MPM. They established a patient-focused research platform named Mesothelioma evolution: deciphering drugable Somatic alterations (MEDUSA), to infer a human evolutionary model of MPM tumorigenesis and understand the inter-relationships that exist between genomic ITH, host immune surveillance and clinical phenotype. It was hoped that their model would also provide a platform for deciphering preclinical models to underpin pharmacogenomic investigations.

They used MEDUSA to interrogate genomic big data, from databases such as dbSNP, COSMIC, and KEGG. They conducted multi-regional whole-exosome sequencing of 90 tumour regions from 22 patients (the MEDUSA22 cohort) undergoing routine surgery by pleurectomy decortication. Matched DNA isolated from whole blood served as a germline reference for calling somatic events. The researchers sampled 4 – 5 regions per patient, which were then anatomically stereotyped across the entire cohort, comprising the apex, pericardium, anterior and posterior costophrenic recesses and the oblique fissure.

Findings using MEDUSA

This study revealed extensive exomic variation between and within patients within. MEDUSA revealed a prognostic cluster, C5, which has the most complex evolutionary trajectory and was associated with shorter survival. In the MEDUSA22 cohort, C5 was found to be entirely epithelioid, which is consistent with a recent study that reported the use of deep learning to identify poor prognostic epithelioid MPMs. Loss of chromosome 4, 3p21 and 9p21.3 were consistently found to be early clonal events, which are most likely involved in initiation.

The tumour suppressor gene, BAP1, is encoded by 3p21. It encodes nuclear deubiquitinase, which regulates histone H2A to regulate transcriptional repression. A BAP1 mutation or 3p21 loss was found to be predominantly heterozygous, with evidence of clonal and subclonal second hits leading to its complete inactivation. Moreover, BAP1 has been reported to supress the polycomb repressive complex PRC2, and its inactivation results in upregulated EZH2 mediated trimthylation of histone H3. Therefore, targeting the PRC12 catalytic core (EZH2) with specific small molecule inhibitors could be used as a therapy for MPM.

In addition to that, the researchers found NF2/-22q to be a predominantly late clonal event, suggesting an early evolutionary constraint. NF2 mutation was found to be positively selected during both early and late evolution. Parallel evolution involving NF2/-22 suggests a deterministic trajectory in MPM that involves Hippo pathway inactivation, as a key bottleneck during progression of mesothelioma. The Hippo pathway is currently being explored as a target for therapy, with molecules being investigated that could restore Hippo pathway mediated tumour suppression to treat MPM.


This study developed the MEDUSA to infer a human evolutionary model of MPM tumorigenesis and understand the inter-relationships that exist between genomic ITH, host immune surveillance and clinical phenotype. Their results suggest that clonal architecture modulated constitutive immune surveillance, which may highlight a potential mechanism leading to de novo or acquired immune checkpoint inhibitor resistance in MPM. To test this, the researchers are conducting clinicogenomic correlative studies in clinical trials involving cohorts from phases II and III.

Image credit: tracerjuice – FreePik

More on these topics

AI / Cancer / Drug Development

Share this article