Methods For Determining Disease Risk Combining Downsampling of Class-Imbalanced Sets With Survival Analysis

Patent metadata

Field	Value
Publication number	US-2022015714-A1
Application number	US-201917297669-A
Country	US
Kind code	A1
Filing date	Nov 21, 2019
Priority date	Nov 29, 2018
Publication date	Jan 20, 2022
Grant date	—

Abstract

A method for downsampling class-imbalanced sets with survival analysis comprising: acquiring a class-imbalanced data set, wherein the class-imbalanced data set comprises biological data from a plurality of subjects, wherein the biological data of each subject includes an observation, a time value, and a plurality of clinical measurements, and wherein the biological data is categorized as being part of a majority data class or a minority data class, wherein the majority data class has a greater number of observations than the minority data class; downsampling the class-imbalanced data set, wherein the downsampling results in the majority data class having an equivalent or substantially equivalent number of observations as the minority data class; and performing cross-validation on the downsampled data set with a survival analysis to generate a survival model, wherein the observation comprises an event or no event at a specific time value.

First claim

1 . A method comprising:
a) acquiring a class-imbalanced data set, wherein the class-imbalanced data set comprises biological data from a plurality of subjects, wherein the biological data of each subject includes an observation, a time value, and a plurality of clinical measurements, and wherein the biological data is categorized as being part of a majority data class or a minority data class, wherein the majority data class has a greater number of observations than the minority data class; b) downsampling the class-imbalanced data set to generate a downsampled data set, wherein the downsampling results in the majority data class having an equivalent or substantially equivalent number of observations as the minority data class; and c) performing cross-validation on the downsampled data set with a survival analysis to generate a survival model; wherein the observation comprises an event or no event at a specific time value, wherein an AUC, sensitivity, specificity, and/or C-index of the survival model is closer to 1 than a AUC, sensitivity, specificity, and/or C-index of a survival model where the class-imbalanced data set was not downsampled prior to the survival analysis.

2 . (canceled)

3 . The method of claim 1 , wherein the class-imbalanced data set is a survival data set.

4 . The method of claim 1 , wherein the event is a disease, disorder, or condition of a subject.

5 . The method of claim 1 , wherein the survival analysis is selected from the group consisting of a Cox proportional hazard analysis, a random forest analysis, an accelerated failure time analysis, and any combination thereof.

6 . The method of claim 5 , further comprising an elastic net penalty.

7 . The method of claim 1 , wherein the cross-validation is at least a 2-fold, 3-fold, 4-fold, 5-fold, 6-fold, 7-fold, 8-fold, 9-fold, 10-fold, 11-fold, 12-fold, 13-fold, 14-fold, 15-fold, 16-fold, 17-fold, 18-fold, 19-fold, or 20-fold cross-validation.

8 . The method of claim 1 , wherein the survival model comprises from 5 to 1,000 features, wherein each feature is selected from the group consisting of a protein measurement, a clinical factor, and a combination thereof.

9 . The method of claim 8 , wherein the clinical factor is selected from the group consisting of age, weight, blood pressure, height, BMI, cholesterol, sex, and a combination thereof.

10 . The method of claim 1 , wherein the clinical measurements are selected from proteomic measurements, genomic measurements, transcriptome measurements, metabolomics measurements, or a combination thereof.

11 . The method of claim 1 , wherein the cross-validation is selected from k-fold cross-validation, Monte Carlo cross-validation, and Leave N Out validation.

12 . The method of claim 1 , wherein the majority data class is 95% of the class-imbalanced data set and the minority data class is 5% of the class-imbalance data set.

13 . The method of claim 1 , wherein the majority data class is 90% of the class-imbalanced data set and the minority data class is 10% of the class-imbalance data set.

14 . The method of claim 1 , wherein the majority data class is 85% of the class-imbalanced data set and the minority data class is 15% of the class-imbalance data set.

15 . The method of claim 1 , wherein the majority data class is 80% of the class-imbalanced data set and the minority data class is 20% of the class-imbalance data set.

16 . The method of claim 1 , wherein the majority data class is 75% of the class-imbalanced data set and the minority data class is 25% of the class-imbalance data set.

17 . The method of claim 1 , wherein the majority data class is 70% of the class-imbalanced data set and the minority data class is 30% of the class-imbalance data set.

18 . The method of claim 1 , wherein the majority data class is 65% of the class-imbalanced data set and the minority data class is 35% of the class-imbalance data set.

19 . The method of claim 1 , wherein the majority data class is 60% of the class-imbalanced data set and the minority data class is 40% of the class-imbalance data set.

20 - 32 . (canceled)

Assignees

• Somalogic Inc

Inventors

• Hagar Yolanda
• Datta Gargi
• Alexander Leigh E
• Hinterberg Michael A

Classifications

• A61B5/05

  Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves (measuring movement of the entire body or parts thereof A61B5/11; detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof A61B5/24) · CPC title

• G16H50/30 Primary

  for calculating health indices; for individual health risk assessment · CPC title

• A61B5/4866

  Evaluating metabolism (using breath test A61B5/083) · CPC title

• A61B5/117

  Identification of persons (methods or arrangements for recognising patterns, e.g. fingerprints, G06F18/00, G06V40/00; identification of persons by analysing their voice or speech G10L17/00) · CPC title

• A61B5/7264

  Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems · CPC title