Ferraro, Luigi (2024) Enhancing biological insight of cancer exploiting the complexity of multi-omics data. [Tesi di dottorato]

Preview

Text Ferraro_Luigi_36.pdf Download (11MB) | Preview

Collection description

Biological systems are intricate, interconnected entities, and their behaviors emerge and evolve dynamically through diverse mechanisms within various internal molecular domains. This dynamic interplay gives rise to complex, multi-layered structures, continually requiring innovative, comprehensive, and sophisticated approaches for thorough exploration. The recent advances in omics sequencing technologies, spanning genomics, transcriptomics, proteomics, and more, has played a crucial role in advancing comprehensive systems biology models, particularly in complex diseases such as cancer. The emergence of multi-omics models has been particularly impactful, demonstrating a significant capability to capture the dynamic interactions within and between various omics layers. These approaches have proven their worth over the years in understanding the molecular dysregulation characterizing tumor systems, with the ultimate goal of translating this knowledge into clinical practices. In this context, a critical area of investigation revolves around characterizing the therapeutic effects of compounds. This field is inherently complex, marked by prolonged timelines from inception to market launch and a notable failure rate. To address these challenges, computational virtual screening powered by machine learning algorithms has emerged as a promising approach for predicting therapeutic efficacy. However, the intricate relationships between the features learned by these algorithms can be exceptionally challenging to decipher. Here we have engineered an artificial neural network model designed specifically for predicting drug sensitivity. This model leverages a biologically informed Visible Neural Network (VNN), enhancing its interpretability. The trained model allows for an in-depth exploration of the biological pathways integral to prediction and the chemical attributes of drugs that impact sensitivity. Our model harnesses multi-omics data derived from an different tumor tissue sources, as well as molecular descriptors that encapsulate the properties of drugs. To establish a robust framework for the versatile integration of bulk and single-cell multi-omics data, we have devised a network-based approach capable of identifying relevant multi-omics relationships. The model generates a latent space of molecular features that accurately identifies and segregates different cancer subtypes or cells. These two tools offer a profound and enlightening perspective on unraveling the molecular intricacies within biological systems. They provide invaluable multi-omics-related resources poised for future applications in both biological research and clinical settings. Our integration tool significantly augments the comprehension of the multifaceted nature of biological data, unlocking insights that are not apparent through a singular omics layer or human curation alone. Incorporating this innovative model into drug development processes carries the potential to advance precision medicine significantly. It promises a deeper understanding of why certain cancer therapies prove effective, potentially revolutionizing treatment strategies.

Downloads

Actions (login required)

View Item