Deep learning in molecular biology marker recognition of patients with acute myeloid leukemia

• Published in: The Journal of supercomputing Vol. 78; no. 9; pp. 11283 - 11297
• Main Authors: Chen, Lieguang, Lu, Ying, Pei, Renzhi, Zhang, Pisheng, Liu, Xuhui, Du, Xiaohong, Chen, Dong, Cao, Junjie, Li, Shuangyue, Zhuang, Xianxu
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2022; Springer Nature B.V
• Subjects: Abnormalities; Algorithms; Belief networks; Biology; Biomedical Image Analysis Using Supercomputing Deep Learning; Compilers; Computer Science; Deep learning; Gene expression; Interpreters; Leukemia; Machine learning; Markers; Molecular biology; Processor Architectures; Programming Languages; Deep belief network; Gene expression level; Acute myeloid leukemia; Wilm’s tumor 1
• DOI: 10.1007/s11227-021-04104-9
• ISSN: 0920-8542

In this study, the deep belief network (DBN) algorithm was used to identify the Wilm’s tumor 1 (WT1) gene expression levels, and then, the role of WT1 expression in the classification of acute myeloid leukemia (AML) was explored. 121 AML patients diagnosed in the hospital from 2017.10 to 2019.10 were selected as the research subjects and set as the AML group. Another 9 non-leukemia patients were selected as the control group. The expression levels of WT1 in the two groups were compared, and DBN was used to classify the patients based on the WT1 expression levels. The real-time quantitative PCR was used to detect the abnormalities of FLT3, PML-RAR, and other molecular markers at different WT1 expression levels. The results showed that the expression of WT1 in AML patients was significantly higher than that in non-leukemia patients. The expression of WT1 in patients of M3 type was the highest, and that was the lowest in patients of the M5 type. The accuracy, precision, recall, and F1 indexes for WT1 expression identification using deep belief network were 94.06%, 93.82%, 93.59%, and 93.63%, respectively. In conclusion, deep learning technology is very sensitive in identifying the molecular biology markers in AML patients, which provides a reference for efficient and intelligent disease diagnosis.