ASCO-SITC 2020 — ER+/HER2- breast cancer: machine learning-assisted model identifies prognostic genes


  • Pavankumar Kamat
  • Univadis
Access to the full content of this site is available only to registered healthcare professionals. Access to the full content of this site is available only to registered healthcare professionals.

Takeaway

  • A machine learning-assisted mathematical model identified genes in the tumor microenvironment (TME) that are strongly associated with prognosis in patients with stage III, estrogen receptor-positive (ER+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer.

Why this matters

  • Stroma in the TME is known to affect prognosis and therapeutic response; however, there are only a few prognostic mathematical models based on mRNA expressivity in the TME.

Study design

  • Using 50 cycles of machine learning, the model categorized 98 patients with ER+, HER- breast cancer from the Cancer Genome Atlas Program into high- and low-risk groups based on mRNA expression of 26 gene groups.
  • The gene groups comprised 191 genes enriched in cellular and noncellular elements of the TME.
  • Funding: None disclosed.

Key results

  • 15 genes, namely CD8A, CD8B, FCRL3, GZMK, CD3E, CCL5, TP53, ICAM3, CD247, IFNG, IFNGR1, ICAM4, SHH, HLA-DOB, and CXCR3 were associated with good prognosis.
  • 5 genes, namely LOXL2, PHEX, ACTA2, MEGF9, and TNFSF4 were associated with poor prognosis.
  • There was a significant survival difference between the high- and low-risk groups (HR, 2.878; P=.05).
  • The high-risk group had a higher expression of genes associated with desmoplastic reaction, neutrophils, and immunosuppressive cytokines, whereas the low-risk group had a higher expression of genes associated with immune system activation (P<.05>

Limitations

  • Other types of breast cancer were not assessed.