Ovarian recurrence risk assessment using machine learning, clinical information, and serum protein levels to predict survival in high grade ovarian cancer

David P Mysona; Sharad Purohit; Katherine P Richardson; Jessa Suhner; Bogna Brzezinska; Bunja Rungruang; Diane Hopkins; Gregory Bearden; Robert Higgins; Marian Johnson; Khaled Bin Satter; Richard McIndoe; Sharad Ghamande

doi:10.1038/s41598-023-47983-z

Ovarian recurrence risk assessment using machine learning, clinical information, and serum protein levels to predict survival in high grade ovarian cancer

Sci Rep. 2023 Nov 27;13(1):20933. doi: 10.1038/s41598-023-47983-z.

Authors

David P Mysona^{1

2}, Sharad Purohit^{3

4

5}, Katherine P Richardson³, Jessa Suhner⁴, Bogna Brzezinska⁴, Bunja Rungruang⁴, Diane Hopkins³, Gregory Bearden³, Robert Higgins⁴, Marian Johnson⁴, Khaled Bin Satter³, Richard McIndoe^{3

4}, Sharad Ghamande⁴

Affiliations

¹ Center for Biotechnology and Genomic Medicine, Medical College of Georgia, Augusta University, 1120 15th Street, Augusta, GA, 30912, USA. dmysona@augusta.edu.
² Department of Obstetrics and Gynecology, Medical College of Georgia, Augusta University, 1120 15th Street, Augusta, GA, 30912, USA. dmysona@augusta.edu.
³ Center for Biotechnology and Genomic Medicine, Medical College of Georgia, Augusta University, 1120 15th Street, Augusta, GA, 30912, USA.
⁴ Department of Obstetrics and Gynecology, Medical College of Georgia, Augusta University, 1120 15th Street, Augusta, GA, 30912, USA.
⁵ Department of Undergraduate Health Professionals, College of Allied Health Sciences, Augusta University, 1120 15th Street, Augusta, GA, 30912, USA.

Abstract

In ovarian cancer, there is no current method to accurately predict recurrence after a complete response to chemotherapy. Here, we develop a machine learning risk score using serum proteomics for the prediction of early recurrence of ovarian cancer after initial treatment. The developed risk score was validated in an independent cohort with serum collected prospectively during the remission period. In the discovery cohort, patients scored as low-risk had a median time to recurrence (TTR) that was not reached at 10 years compared to 10.5 months (HR 4.66, p < 0.001) in high-risk patients. In the validation cohort, low-risk patients had a median TTR which was not reached compared to 4.7 months in high-risk patients (HR 4.67, p = 0.009). In advanced-stage patients with a CA125 < 10, low-risk patients had a median TTR of 68 months compared to 6 months in high-risk patients (HR 2.91, p = 0.02). The developed risk score was capable of distinguishing the duration of remission in ovarian cancer patients. This score may help guide maintenance therapy and develop innovative treatments in patients at risk at high-risk of recurrence.

MeSH terms

Blood Proteins
Female
Humans
Machine Learning
Neoplasm Recurrence, Local
Ovarian Neoplasms* / drug therapy
Risk Assessment
Risk Factors

Substances

Blood Proteins