Classifiers for detection of endometriosis

Abstract

Described herein are improved methods for the detection of endometriosis. Generally, the methods include, but are not limited to, applying machine learning algorithm to miRNA levels in order to detect, predict, diagnose, or monitor the presence or absence of endometriosis.

1 . A method of detecting and treating endometriosis or a non-endometriosis condition in a female subject, comprising:
 (a) detecting in a bodily fluid sample from the female subject an expression profile of a panel of miRNAs associated with endometriosis, wherein the panel of miRNAs associated with endometriosis comprises miR-342 or miR-451a;   (b) applying a machine learning algorithm to the expression profile of the panel of miRNAs associated with endometriosis, wherein the machine learning algorithm has importance measures assigned to miRNA features, and wherein:   i. an importance measure is assigned to miR-342 and the importance measure assigned to miR-342 is greater than the importance measure assigned to miR-150, miR-3613, miR-451a, let-7b, or miR-125b; or   ii. an importance measure is assigned to miR-451a and the importance measure assigned to miR-451a is greater than the importance measure assigned to miR-3613, miR-125b, or let-7b;   (c) using the machine learning algorithm to detect endometriosis or the non-endometriosis condition in the female subject; and   (d) treating the endometriosis or non-endometriosis condition detected in the female subject with a treatment for endometriosis or with a treatment for a non-endometriosis condition, respectively.   
     
     
         2 . The method of  claim 1 , wherein the importance measure assigned to miR-342 is greater than the importance measure assigned to miR-150, miR-3613, miR-451a, let-7b, or miR-125b. 
     
     
         3 . The method of  claim 1 , wherein the importance measure assigned to miR-451 a is greater than the importance measure assigned to miR-3613, miR-125b or let-7b. 
     
     
         4 . The method of  claim 1 , wherein the importance measure assigned to miR-342 is greater than the importance measure assigned to at least two of: miR-150, miR-3613, miR-451a, let-7b, and miR-125b. 
     
     
         5 . The method of  claim 1 , wherein the bodily fluid sample is a cell-free sample. 
     
     
         6 . The method of  claim 1 , wherein the bodily fluid sample is a blood sample, a plasma sample, a saliva sample, or a serum sample. 
     
     
         7 . The method of  claim 1 , wherein applying a machine learning algorithm to the expression profile comprises applying a machine learning algorithm with specific importance measure rankings assigned to the miRNA features, wherein the specific importance measure rankings from highest to lowest is miR-342, miR-451a, miR-3613, miR-125b, let-7b, and miR-150. 
     
     
         8 . The method of  claim 1 , wherein the machine learning algorithm is a random forest algorithm, k-nearest-neighbors algorithm (KNN), support vector machine (SVM), or Naive Bayes. 
     
     
         9 . The method of  claim 1 , wherein the machine learning algorithm is a random forest algorithm. 
     
     
         10 . The method of  claim 1 , wherein the method detects endometriosis in a population of women with a specificity of greater than 80%. 
     
     
         11 . The method of  claim 10 , wherein the population of women is premenopausal women. 
     
     
         12 . The method of  claim 10 , wherein the population of women comprises women with leiomyomas, cystadenomas, chronic pelvic infections, teratomas, endometriomas, or paratubal cysts. 
     
     
         13 . The method of  claim 10 , wherein the population of women comprises women with Stage I/II endometriosis. 
     
     
         14 . The method of  claim 10 , wherein the population of women comprises women with Stage III/IV endometriosis or women with any stage of endometriosis 
     
     
         15 . The method of  claim 10 , wherein the population of women comprises women having received hormone therapy within 3 months of the date on which the bodily fluid sample was obtained or women at any phase of their menstrual cycle. 
     
     
         16 . The method of  claim 10 , wherein the population of women comprises a cohort comprising at least 100 women. 
     
     
         17 . The method of  claim 1 , wherein the machine learning algorithm is trained on expression data from at least 100 samples. 
     
     
         18 . The method of  claim 1 , wherein the machine learning algorithm is trained on a population of women comprising women having any stage of endometriosis. 
     
     
         19 . The method of  claim 1 , wherein the method has an AUC for detecting endometriosis of greater than 0.85 in a population of women. 
     
     
         20 . The method of  claim 1 , wherein the treatment for the non-endometriosis condition does not comprise surgery.