Systems and methods of diagnosing idiopathic pulmonary fibrosis on transbronchial biopsies using machine learning and high dimensional transcriptional data

Abstract

The present invention provides systems, methods, and classifiers for differentiating between samples as usual interstitial pneumonia (UTP) or non-UIP.

1 .- 30 . (canceled) 
     
     
         31 . A method for determining that a subject is positive for a non-usual interstitial pneumonia (non-UIP) subtype of a plurality of non-UIP subtypes, comprising:
 (a) obtaining a biological sample of said subject;   (b) assaying nucleic acid molecules derived from said biological sample to identify a level of expression of at least one gene associated with said non-UIP subtype; and   (c) processing said level of expression to generate a classification of said biological sample as being positive for said non-UIP subtype.   
     
     
         32 . The method of  claim 31 , wherein said non-UIP subtype is hypersensitivity pneumonitis (HP), non-specific interstitial pneumonia (NSIP), sarcoidosis, respiratory bronchiolitis (RB), bronchiolitis, diffuse alveolar damage (DAD) or organizing pneumonia (OP). 
     
     
         33 . The method of  claim 31 , wherein said biological sample is a transbronchial biopsy sample or a bronchoalveolar lavage sample. 
     
     
         34 . The method of  claim 31 , wherein (b) comprises sequencing. 
     
     
         35 . The method of  claim 31 , wherein said assaying further comprises identifying a level of expression of at least one control nucleic acid molecule in said biological sample. 
     
     
         36 . The method of  claim 31 , wherein said plurality of non-UIP subtypes comprise hypersensitivity pneumonitis (HP), non-specific interstitial pneumonia (NSIP), sarcoidosis, respiratory bronchiolitis (RB), bronchiolitis, diffuse alveolar damage (DAD) or organizing pneumonia (OP). 
     
     
         37 . The method of  claim 31 , wherein (c) is performed using a machine learning algorithm that is trained to identify said non-UIP subtype of said plurality of non-UIP subtypes. 
     
     
         38 . The method of  claim 37 , wherein said machine learning algorithm is trained using features comprising gene expression variants, gene fusions, loss of heterozygosity, or biological pathway effect. 
     
     
         39 . The method of  claim 38 , wherein said gene expression variants are alternative splice variants. 
     
     
         40 . The method of  claim 37 , wherein said machine learning algorithm is trained with a training set that is independent of said biological sample. 
     
     
         41 . The method of  claim 31 , wherein said biological sample is fresh-frozen or fixed. 
     
     
         42 . The method of  claim 31 , wherein said nucleic acid molecules are ribonucleic acids (RNA) molecules, and wherein said assaying comprises generating complementary deoxyribonucleic acid (cDNA) molecules from said RNA molecules. 
     
     
         43 . The method of  claim 31 , wherein said subject is suspected of having an interstitial lung disease based at least in part on one or more clinical signs or one or more symptoms. 
     
     
         44 . The method of  claim 44 , wherein said one or more symptoms comprise shortness of breath or dry cough. 
     
     
         45 . The method of  claim 44 , wherein said on or more clinical signs comprise a result of an imaging test, a pulmonary function test, or a lung tissue analysis. 
     
     
         46 . The method of  claim 45 , wherein said imaging test is chest X-ray or computerized tomography. 
     
     
         47 . The method of  claim 46 , wherein said computerized tomography is high-resolution computerized tomography. 
     
     
         48 . The method of  claim 45 , wherein said pulmonary function test is spirometry, oximetry, or an exercise stress test. 
     
     
         49 . The method of  claim 45 , wherein said lung tissue analysis comprises histological or cytological analysis of a lung tissue sample of said subject. 
     
     
         50 . The method of  claim 31 , further comprising providing a therapeutic intervention to said subject based at least in part on said classification generated in (c).