Combining Molecular Biomarkers for Cancer Diagnosis
Author Information
Author(s): Mamtani Manju R, Thakre Tushar P, Kalkonde Mrunal Y, Amin Manik A, Kalkonde Yogeshwar V, Amin Amit P, Kulkarni Hemant
Primary Institution: Lata Medical Research Foundation, Nagpur, India
Hypothesis
Can a statistical algorithm improve the diagnostic performance of multiple molecular biomarkers for cancer classification?
Conclusion
The proposed algorithm can accurately diagnose different cancer states using a combination of biomarkers.
Supporting Evidence
- The algorithm achieved 100% diagnostic accuracy in the OvCa dataset.
- Diagnostic performance was validated across multiple datasets.
- The method is computationally less intense than other data mining techniques.
Takeaway
This study created a smart way to use different cancer markers together to help doctors tell if someone has cancer.
Methodology
The algorithm involves estimating the area under the ROC curve for each biomarker, selecting a subset using linear regression, and combining them with linear discriminant analysis.
Potential Biases
Potential bias in selecting training sets may affect the generalizability of results.
Limitations
The algorithm is only validated for dichotomous classification and may not identify all differentially expressed biomarkers.
Participant Demographics
The study involved datasets from various cancer types including ovarian cancer, lung cancer, and breast cancer.
Statistical Information
Confidence Interval
95% confidence interval for the LuMe dataset was 99.7%–100%.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website