Machine Learning Algorithms for Breast Cancer Prediction

Main Article Content

Senthil Kumar K M.E
A. Akalya
J. L. Gayathri
V. Kanimozhi

Keywords

Semi-supervised learning, deep learning, genomics, multi-omics, and variation auto-encoding

Abstract

There are numerous subtypes of breast cancer, each with its own unique outlook. The evaluation of the expression of small gene sets is the primary focus of the current stratification methods. In the upcoming years, Next Generation Sequencing (NGS) is anticipated to generate a significant amount of genomic data. We investigate the application of deep learning, or machine learning, to the subtyping of breast cancer in this case study. We used pan-cancer and non-cancer data to create semi-supervised settings because there weren't any publicly accessible data. A wide range of supervised and semi-supervised designs are investigated with the help of Integrative omics data like microRNA expression and copy number variations.On our gene expression data challenge, accuracy results indicate that simpler models perform better than deep semi-supervised approaches. Deep model performance improves only marginally (if at all) when integrated combining several omics data types emphasises the need for additional research on bigger datasets of multi-omics data as they become accessible. In terms of biology, our linear model typically confirms. earlier classifications of gene subtypes. The development of a more varied and unexplored set of representative omics traits that may be helpful for subtyping breast cancer has resulted from deep methods, which imitate non-linear interactions.

Abstract 161 | pdf Downloads 134

References

1. T. Srlie et al., "Gene expression patterns of breast carcinomas define tumor subtypes with clinical implications," Proc. Nat. Acad. Sci. United States, vol. 98, no. 19, pp. 10869- 10874, 2001.
2. F. Vieira and F. Vieira Schmitt, "An update on multigene predictive testing for breast cancer-emergent clinical signs," Front. Med., vol. 5, no. 248, 2018.
3. J. Parker et al., "Supervised risk predictor of breast cancer based on intrinsic subtypes," J. Clin. Oncology, vol. 27, pp. 1160-1167, 2009.
4. B. Wallden et al., "Development and validation of the PAM50-based prosigna breast cancer gene signature test," BMC Med. Vol. 1 Genomics Art. no. 54, 8, no. 1, 2015.
5. cancer Genome Atlas Network et al., Nature, vol. 490, no. 7418 (2012), pp. 61-70.