Molecular Classifiers for Accurate Prostate Cancer Risk Grading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for distinguishing between low and high-grade prostate cancer using needle biopsies are inaccurate, leading to inappropriate treatment decisions due to sampling error and inter-observer variability, placing patients in the wrong risk category and potentially causing undue morbidity or missing aggressive cancer treatments.
Innovation Solution
Development of molecular classifiers, PRONTO-e and PRONTO-m, which integrate mRNA, CNA, methylation, and clinical features to predict disease progression risk by comparing patient features to a trained classifier using machine learning algorithms, improving the accuracy of prostate cancer grade assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If needle biopsy is used for prostate cancer diagnosis, then minimally invasive sampling is achieved, but measurement precision of cancer grade is poor due to sampling error and inter-observer variability
Solution Approach 1:
The patent replaces mechanical/pathological grading methods with molecular profiling using RNA and DNA analysis. Instead of relying on visual assessment of glandular architecture by pathologists (mechanical observation), the system uses molecular classifiers that process genetic and transcriptional data to predict cancer grade, thereby achieving higher measurement precision while maintaining the minimally invasive nature of needle biopsy sampling.
Solution Approach 2:
The patent transforms the diagnostic parameters from morphological features (glandular architecture, cellular morphology) to molecular parameters (gene expression profiles, copy number variations, methylation patterns). This parameter transformation enables more precise and objective cancer grade assessment by measuring fundamental molecular characteristics that are less susceptible to sampling error and observer variability.
2Measurement precision
If molecular classifiers are developed to improve cancer grade assessment, then measurement precision improves, but device complexity increases due to multiple data types and machine learning algorithms
Solution Approach 1:
The patent segments the molecular classifier into distinct functional modules: (1) data processing module for handling multiple data types (RNA, DNA, methylation), (2) feature selection module for identifying relevant molecular markers, (3) machine learning classification module for predicting cancer grade, and (4) interpretation module for clinical application. This segmentation allows the complex system to be managed through modular components, each performing a specific function.
Solution Approach 2:
The patent creates a universal molecular profiling platform that can handle multiple types of molecular data (transcriptomic, genomic, epigenomic) through a single integrated classifier system. The machine learning model is designed to process diverse data types and provide unified cancer grade predictions, making the system multi-functional and reducing the need for separate diagnostic tools for different molecular analyses.
3Reliability
If comprehensive molecular profiling is performed, then reliability of cancer grade prediction improves, but loss of time increases due to extensive testing and data analysis
Solution Approach 1:
The patent performs preliminary actions by pre-selecting and pre-processing the most relevant molecular features and training robust machine learning models in advance. The classifier system includes pre-identified gene panels and processed reference data that can be quickly applied to new samples, reducing the time needed for analysis while maintaining high reliability through the pre-optimized predictive model.
Solution Approach 2:
The patent applies partial action by focusing molecular profiling on the most informative features rather than analyzing every possible molecular parameter. The machine learning model is trained to identify and weight the most critical features, allowing the system to achieve high prediction reliability by concentrating on key molecular drivers rather than exhaustively analyzing all molecular data, thereby reducing analysis time.
Data Source
Figure 1A~1C
Figure 2
Figure 3A~3B
AI summary
There is described herein a method of predicting disease progression risk in a subject with prostate cancer, the method comprising: a) providing a sample containing RNA and DNA material from tumour cells; b) determining or measuring values for substantially all of patient features listed for PRONTO-e or PRONTO-m in Table 6, and some or all reference or control features set forth in Table 6; c) comparing said patient features to the reference or control features; and d) computing a prediction score using a classifier that takes said patient feature values as input, the classifier having been previously trained on samples from a population of early prostate cancer patients.