Developing diagnostic and prognostic protein signatures for ovarian cancer
Advancing gynecological health with breakthrough protein signatures for early detection and improved prognosis using Olink’s next-generation proteomics technology.
Ovarian cancer often eludes early detection with non-specific symptoms, leading to unfavorable long-term outcomes. To improve survival rates, there is an urgent need for methods that can detect the disease earlier with improved accuracy. Leveraging the capabilities of Olink’s scalable PEA technology, scientists from Uppsala University led by Professor Ulf Gyllensten embarked on a project to identify and establish protein signatures for ovarian cancer.
Biomarker discovery identifies an 11-protein diagnostic signature for ovarian cancer
In the initial biomarker discovery phase, approximately 600 plasma proteins were analyzed in three cohorts of over 400 cancer patients and controls using multiple Olink Target 96 panels. This identified several proteins that were differentially expressed between cases and controls, with significant associations to ovarian cancer after multivariate analysis. Modeling of this data identified an 11-protein signature capable of discerning ovarian cancer (stages I-IV) from healthy control samples with an impressive accuracy of around 94%.
We have shown that the performance of the risk score model is robust across cohorts. Also, the risk score pattern after diagnosis follows the common clinical responses during treatment and relapse/progression and may be useful in monitoring clinical developments during follow-up.
Development and validation of the diagnostic signature
While the ovarian cancer signature identified in the proteomic discovery phase (comprising, MUCIN-16, SPINT1, TACSTD2, CLEC6A, ICOSLG, MSMB, PROK1, CDH3, WFDC2, KRT19 and FR-alpha) showed impressive diagnostic performance, any future utility in the clinic would require development of a custom panel based on these proteins and its validation in independent patient cohorts. Consequently, they developed a custom panel together with Olink that encompassed the 11 identified proteins, and then meticulously validated this panel using an additional cohort.
Notably, the custom Olink Focus panels provide a significant advantage for clinical applications by allowing readouts in standard concentration units (pg/mL) in addition to the relative quantification (NPX units) provided by higher multiplexed PEA panels.
Risk score for healthy women and those with benign tumors, or endometrial, cervical and ovarian cancer (stages I–IV). The dotted line indicates the cut-off for malignancy at the ‘best-point’ established for ovarian cancer. The numbers above each group represent the number of samples and the percent of samples that are above the cut-off. Reproduced from Enroth et al. (2022) Communications Medicine, DOI: 10.1038/s43856-022-00193-6.
A large-scale clinical validation study was conducted, involving two independent cohorts comprising over 1,100 clinical samples. The results of this thorough validation study confirmed the effectiveness of the previously identified multi-protein model.
In these independent cohorts, the model demonstrated a sensitivity of 0.83-0.91 and specificity of 0.88-0.92 for ovarian cancer vs controls, further establishing its robustness. It also showed good specificity in discriminating ovarian cancer from other gynecological malignancies.
A diagnostic signature with prognostic potential
Following on from the validation of the diagnostic performance of the signature, the team used a data-driven modeling approach to analyze risk score patterns during a two-year follow-up period after diagnosis, revealing four distinct risk score trajectories that correlated with clinical progression and survival outcomes. Notably, analysis focusing on 5-year survival indicated that the risk score at the time of diagnosis was the second most powerful predictor of survival, surpassed only by tumor stage.
