A machine learning and multi-omics approach to predict measurable residual disease response and understand resistance in multiple myeloma

Abstract

The doctoral thesis by Camila Guerrero delves into the multifaceted nature of multiple myeloma (MM), a complex haematological malignancy characterised by the neoplastic proliferation of clonal plasma cells. The thesis aims to enhance the understanding of MM's pathobiology, therapeutic responses, and resistance mechanisms using a machine learning and multi-omics approach, integrating clinical, cytogenetic, tumor immune microenvironment, and high-throughput biological data. The results have been published in two prestigious journals: Clinical Cancer Research, and Blood Journal, with an additional manuscript under preparation, and presented in several international and national conferences such as the American Society of Hematology and the International Myeloma Workshop. Multiple myeloma is the second most common haematological cancer, often progressing from precursor states such as monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMM). The disease is defined by the clonal proliferation of plasma cells in the bone marrow, leading to the secretion of monoclonal immunoglobulins that is associated organ dysfunctions. The therapeutic landscape for MM has expanded significantly in the last decade, with the most recent inclusion of monoclonal antibodies to the standard of care that includes proteasome inhibitors and immunomodulatory drugs with or without high dose therapy and autologous stem cell transplant. Measurable residual disease (MRD) has emerged as a critical prognostic marker, and its assessment through advanced techniques like next-generation flow cytometry (NGF) and sequencing (NGS) has increased its sensitivity. Guerrero's research leverages a machine learning and multi-omics approach to dissect the molecular underpinnings of MM and its resistance to treatment, particularly of circulating tumor cells (CTCs) and MRD, two high-risk clones whose detection is associated with worse survival outcomes. Firstly, clinical, immune and cytogenetic data from three clinical trials were integrated to generate a machine learning model to predict which patients were going to have an undetectable MRD assessment after the first-line of therapy. This model is available online at www.MRDpredictor.com. Secondly, Fine-Gray statistical analyses were used to define a model to predict MRD-resurgence and/or progressive disease. This model is dynamic, and integrates an ISS III staging, late achievement of MRD-negativity and a high percentage of CTCs to define which patients are more likely to exhibit MRD-resurgence and/or progress. Finally, by integrating data from genomics and transcriptomics, the thesis provides a holistic view of the biology of the disease. This approach facilitates the identification of novel biomarkers and therapeutic targets, offering insights into the mechanisms of drug resistance and potential ways for overcoming it. High-throughput techniques such as next-generation flow cytometry for immune profiling, bulk RNA sequencing for gene expression analysis, and whole exome sequencing (WES) for mutational profiling were used to generate comprehensive datasets that inform the multi-omics analyses. Finally, the integration of multi-omics data with clinical, immune and cytogenetic data can significantly enhance the precision of risk stratification and the development of personalised therapeutic strategies. Future research directions include further validation of identified biomarkers and the exploration of novel therapeutic combinations to overcome resistance. In summary, Camila Guerrero's thesis provides a detailed examination of multiple myeloma through a multi-omics and biostatistical lens, offering valuable insights into its pathogenesis, treatment responses, and resistance mechanisms. The work underscores the potential of integrated clinical and omics data to transform MM management and improve patient prognoses.