Machine learning deciphers millions of genome samples, unlocks cancer cure | Hfocus.org Health Systems Insights

Machine Learning (ML) or “Machine Learning” It’s part of artificial intelligence, where machine learning algorithms create models based on sampled data, called “training data,” to make unprogrammed predictions or decisions. The computer learns from the information provided to perform certain tasks. for simple tasks assigned to computers that have greatly contributed to saving human effort without people needing to waste time configuring algorithms themselves (1)

machine learning Can help work in a variety of branches even in the medical field And it is advancing rapidly in the field. For example, in 2012, the co-founder of Sun Microsystems, Vinod Khosla, estimated that there would be 80 % of doctor jobs lost in the next two decades of machine learning medical diagnostic software. , machine learning technology will be used to help diagnose and assist researchers in developing a cure for COVID-19 (3).

Latest news articles from the University of Helsinki. published September 6, 2022, Nature Communications. presented a method for accurate genomic data analysis in cancer biopsy. This tool uses machine learning methods. to repair damaged DNA and reveal the actual mutation process in tumor samples. This has helped unlock new approaches to cancer treatment and produce highly valuable therapeutic drugs in millions of archived cancer samples. (4)

Molecular Diagnosis Helps Match Patients to the Right Cancer Treatment Researchers are particularly interested in DNA profiling in clinical cancer samples. Qingli Guo from the University of Helsinki The lead author of this research said: This invaluable cancer resource is currently not used for molecular diagnostics due to poor DNA quality. It also causes severe DNA damage, which is an inevitable challenge when analyzing cancer genomes in preserved tissues.

Analysis of mutations in cancer genomes can help detect cancer in its early stages. Accurately diagnosing cancer and revealing why some cancers are resistant to treatment This new method can greatly speed up the development of clinical applications. This can directly affect the care of cancer patients in the future.

interesting is that the new method can predict the development of cancer processes by more than 90%.

Qingli Guo works closely with scientists from the Institute of Cancer Research (ICR) London and Queen Mary University of London. Develop a machine learning/machine learning method called FFPEsic to study how formalin causes DNA mutations.

The results show that almost half of cancer screening will usually be missed without correcting for confounding factors in the sample, but more than 90% of the FFPEsic was used to predict correct

Cancer develops gradually, profiling the mutation process in horizontal samples. Longitudinal data, tracking the same sample at different time points, can help identify clinical predictors and diagnose disease at different tumor stages.

“Our findings enable the identification of clinically relevant signatures from tumor biopsies that have been stored at room temperature for decades. with a deeper understanding of how formalin affects the cancer genome Our study offers a huge opportunity to transform the form of identity testing developed using large-scale cost-effective archival samples.” Qingli Guo said (4)

This is one of the new developments in the use of machine learning to treat cancer. Progress has been made in the use of machine learning for this purpose, for example, a new deep learning method developed by researchers at the Koch Institute for Integrative Cancer Research at MIT and Massachusetts General Hospital (MGH) was recently announced which could help classify idiopathic cancer by looking at the gene expression programs involved in early cell development and differentiation (5).

Cancer cells look and behave very differently to normal cells. This is partly due to dramatic changes in gene expression. But because of advances in single cell profiling and efforts to catalog different cell expression patterns in cell mapping. There is a lot of information that shows that there are different types of cancer. how did this happen This is a very useful use of data for human life.

However, the challenge of using data and machine learning to help catch cancer is that if the model is too complex and shows too many features of cancer gene expression, the model may appear to have completely learned the training data . but stumbles when new information is found. But if the model is simplified by reducing the number of features The model may lose different types of information that would lead to the correct classification of cancer.

Machine learning is not perfect. But it is ready to be perfected in the future. At least the use of its algorithm has greatly reduced the burden of human diagnostics.

For example, the research team at the Koch Institute has balanced machine learning in the analysis of different types of cancer samples. The researchers compared two large cell maps, identifying the relationship between tumors and embryonic cells: the Cancer Genome Atlas (TCGA), which contains gene expression data for 33 tumor types, and the Mouse Cell Organogenesis Atlas (MOCA ), which Route 56 separates embryonically. cells as they develop and differentiate.

The resulting map of the relationship between developmental gene expression patterns in tumors and embryonic cells was converted into a machine learning model. The researchers isolated the gene expression of tumor samples from the TCGA into individual components. which corresponds to a specific point in time in the development path and assigns a mathematical value to each of these elements. The researchers then created a machine learning model called the Developmental Multilayer Perceptron (D-MLP), which determines the components of tumor development and predicts its origin

After the machine learning system was trained, D-MLP was applied to 52 new samples of difficult-to-diagnose cancers, particularly of unknown primary cancers that could not be diagnosed using existing instruments These cases represent some of the most challenging ones encountered in MGH hospitals over a 4 4 period starting in 2017. The model classifies four tumor types and provides predictive results and other data that can guide the diagnosis and treatment of these patients

refer

1. Ethem Alpaydin (2020). Introduction to Machine Learning (Fourth ed.). MIT. pp. xix, 1–3, 13–18. ISBN 978-0262043793.

2. Vinod Khosla (January 10, 2012). “Do We Need Doctors or Algorithms?”. Technology Crush.

3. Vaishya, Raju; Javaid, Mohd.; Khan, Ibrahim Haleem; Haleem, Abid (July 1, 2020). “Artificial Intelligence (AI) applications for the COVID-19 pandemic”. Diabetes and Metabolic Syndrome: Research and Clinical Reviews. 14(4):337–339. doi:10.1016/j.dsx.2020.04.012. PMC 7195043. PMID 32305024.

4. “Researchers use machine learning to unlock the genomic code in clinical cancer samples”. (6-SEP-2022). HELSINKI UNIVERSITY.

5. Bendta Schroeder. (September 1, 2022). “Using machine learning to identify undiagnosed cancers”. MIT NEWS.

National Cancer Institute photo,

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.