Artificial Intelligence for Oncogenomics: An Overview of Use and Impact
Written by: Mark Titleman
The sophistication of artificial intelligence has become such that the practical decisions and theoretical insights once on the mind of the scientist – not entirely dissimilar from the treatments still penned and carefully pondered by doctors – may soon be arrived at automatically. The accuracy of medical diagnosis has improved, too; machine learning algorithms parse data more minutely, interpreted within a statistical algorithmic framework, for approaching perfect efficiency, and deep learning constructs any necessary theories or rules from classification models. Cancer detection is of absolutely vital importance, but relies on both diagnostic tests and the understanding of and screening for cancer genetics, which has progressed slowly. Artificial intelligence of all types promises advances in the area of cancer genetics (which genes cause cancer and how) and their correlation to treatment options. Such a breakthrough, the long sought-after, seemingly inerasable philosopher’s stone of oncology, would yield enormous benefits for mankind while improving understanding of the workings of cancer and molecular biology and marking a significant milestone for artificial intelligence impact.
Molecular biology is the study of large molecules in cells and is coterminous with biochemistry: the associated processes, conformations and energies of large molecules in cells and requisite laboratory techniques. Computing has been used for many decades to enhance the study of these disciplines, for example the generation of Ramachandran plots which detail energetically permissible protein angles from the major structures formed by constituent amino acids. Biochemistry and molecular biology involve the study of tissues and their cells, and oncology involves identifying cancerous tissues via diagnostic tests such as mammograms and Pap tests, understanding the mechanisms of cancer pathogenesis via molecular and cell biology, and cataloging and studying associated genes via oncogenomics. Artificial intelligence ultimately performs tasks via time-efficient and mutable algorithms, discovering patterns in the data for improved cancer and oncogenomic detection and understanding. Oncogenomic advances in particular have cleared the way for an influx of targeted cancer therapies such as Avastin, Gleevec and Herceptin – advances and therapies that will form the bedrock of future oncology. AI for cancer detection is of major use and somewhat widely employed, but its increasing use for oncogenomics offers the most significant promise in diagnostic and treatment development.
Cancerous tissues are tissues that have lost their ability to regulate growth as cells evade apoptosis or cell death. If oncogenes are over-expressed, tumor suppressor genes are under-expressed, or error-correcting machinery is altered due to genetic changes in the cell – the causes of which are multifactorial – then cells with significant genetic changes can proliferate to form tumors with cells containing highly differentiated genes implementing various modes of survival. The mutations that lead to such changes in cell division and survival are innumerable and thus patterns have been sought in the genomic and pathogenetic data.
Machine learning and deep learning are used to mine information in genomics, transcriptomics and proteomics for a deeper understanding of cancerous genes and their expression. This efficient parsing and classification of big data establishes models from which biomarkers related to diagnosis and treatment outcomes can be generated and predicted, ultimately improving treatment options. This is of immense use as identifying patterns in patient biomarkers is too complex for conventional molecular biology and medically cumbersome. Ideally, these new forms of efficient prediction from multi-omics big data can lead to improved comprehension of cancer causes and pathologies and, in conjunction with longitudinal experimental design, can clarify the relationships between the omics for disease generally and lead to the customized treatment plans and targeted therapies of precision medicine. Analysis of multi-omics data via artificial intelligence has the potential to revolutionize all aspects of cancer therapy: risk stratification, cancer subtyping, prediction and clinical decision-making.
The current state of targeted therapies is poorly developed due to few druggable targets and ineffective population coverage. Medi et al. integrated gene expression with genome-wide molecular networks to identify new therapeutic RNA and protein targets for cervical cancer. Laura et al. developed a biology analysis integrating different layers of genomic information for forms of polygenic expression of cancer, including transcription factor co-targeting and protein-protein interaction, and discovered novel cancer driver genes for pancreatic cancer. New algorithms are always being developed, and techniques such as a variation on machine learning called transfer learning have been implemented to fine-tune models with new data sets. Specific types of AI can provide more novel analyses, such as Pattern Computer by SONAR which applies dimension reduction to identify key relevant features of disease (genes), their associations, and if they are actionable or druggable. This and other research have shown immense progress in the identification of cancerous genes, their actionability, and ultimately the creation and potential creation of targeted therapies in precision medicine.
Cancer detection is of utmost importance to patient prognosis, yet while artificial intelligence promises revolutionary direct detection it can also improve far more: associated diagnostic and in particular treatment predictions based on oncogenomics – a process that would in parallel enhance the efficiency of the associated AI and even offer lessons for wider use. These advances will lead to improved biomarker comprehension via modelling and novel targeted therapies in precision medicine oncology. Powerful analysis of multi-omics data via artificial intelligence has the potential to revolutionize virtually all aspects of cancer therapy as well. AI has thus greatly aided in identifying cancerous genes and will continue to do so, which will significantly benefit patient treatment via advances in diagnosis, prognosis, and drug development.
Bibliography
Liao, Jinzhuang, Xiaoying Li, Yu Gan, Shuangze Han, Pengfei Rong, Wei Wang, Wei Li, and Li Zhou. "Artificial intelligence assists precision medicine in cancer treatment." Frontiers in oncology 12 (2023): 998222.
Mohammed, Irshad, Nidhi Singh, and Meenakshi Venkatasubramanian. "A MACHINE LEARNING APPROACH TO PATTERN DISCOVERY IN ONCOGENOMICS." (2019).
Sinha S, Vegesna R, Mukherjee S, et al. PERCEPTION predicts patient response and resistance to treatment using single-cell transcriptomics of their tumors. Nature Cancer. April 18, 2024. DOI: 10.1038/s43018-024-00756-7.
Strausberg, Robert L., Andrew JG Simpson, Lloyd J. Old, and Gregory J. Riggins. "Oncogenomics and the development of new cancer therapies." Nature 429, no. 6990 (2004): 469-474.
Wei, Lise, Dipesh Niraula, Evan DH Gates, Jie Fu, Yi Luo, Matthew J. Nyflot, Stephen R. Bowen, Issam M. El Naqa, and Sunan Cui. "Artificial intelligence (AI) and machine learning (ML) in precision oncology: a review on enhancing discoverability through multiomics integration." The British Journal of Radiology 96, no. 1150 (2023): 20230211.
You, Yujie, Xin Lai, Yi Pan, Huiru Zheng, Julio Vera, Suran Liu, Senyi Deng, and Le Zhang. "Artificial intelligence in cancer target identification and drug discovery." Signal Transduction and Targeted Therapy 7, no. 1 (2022): 156.