Predictive Test for Aggressiveness or Indolence of Prostate Cancer from Mass Spectrometry of Blood-Based Sample
AssigneeBiodesix, Inc.
InventorsJoanna Röder, Heinrich Röder, Carlos Oliveira
Life of this patent
Estimating the term…
Abstract
A programmed computer functioning as a classifier operates on mass spectral data obtained from a blood-based patient sample to predict indolence or aggressiveness of prostate cancer. Methods of generating the classifier and conducting a test on a blood-based sample from a prostate cancer patient using the classifier are described.
Claims15
A method for predicting the aggressiveness or indolence of prostate cancer in a patient previously diagnosed with prostate cancer, comprising the steps of:
- obtaining a blood-based sample from the prostate cancer patient;
- conducting mass spectrometry of the blood-based sample with a mass spectrometer and thereby obtaining mass spectral data including intensity values at a multitude of m/z features in a spectrum produced by the mass spectrometer;
- performing pre-processing operations on the mass spectral data;
- classifying the sample with a programmed computer implementing a classifier operating on the intensity values of the sample after the pre-processing operations are performed and a set of stored intensity values of m/z features from a constitutive set of mass spectra obtained from blood-based samples of a multitude of prostate cancer patients;
- wherein the classifier produces a class label for the blood based sample of High, Early, or the equivalent signifying the patient is at high risk of early progression/relapse of the prostate cancer indicating aggressiveness of the prostate cancer, or Low, Late or the equivalent, signifying that the patient is at low risk of early progression/relapse of the prostate cancer indicating indolence of the prostate cancer.
A system for prostate cancer aggressiveness or indolence prediction, comprising:
- a computer system including a memory storing a final classifier in memory defined from one or more master classifiers generated by conducting logistic regression with extreme drop-out on a multitude of mini-classifiers which meet predefined filtering criteria, a set of mass spectrometry feature values for a constitutive set for classification, the set of mass spectrometry feature values obtained from blood-based samples of prostate cancer patients, a classification algorithm and a set of logistic regression weighting coefficients derived from a combination of filtered mini-classifiers with dropout regularization;
- the computer system including program code for executing the final classifier on a set of mass spectrometry feature values obtained from mass spectrometry of a blood-based sample of a human with prostate cancer.
A laboratory test system for conducting a test on a blood-based sample from a prostate cancer patient to predict aggressiveness or indolence of the prostate cancer comprising, in combination:
- a mass spectrometer conducting mass spectrometry of the blood-based sample with a mass spectrometer and thereby obtaining mass spectral data including intensity values at a multitude of m/z features in a spectrum produced by the mass spectrometer;
- and
- a programmed computer including code for performing pre-processing operations on the mass spectral data and classifying the sample with a final classifier defined by one or more master classifiers generated as a combination of filtered mini-classifiers with regularization, the final classifier operating on the intensity values of the sample after the pre-processing operations are performed and a set of stored values of m/z features from a constitutive set of mass spectra obtained from blood-based samples of prostate cancer patients;
- the programmed computer producing a class label for the blood-based sample of High, Early or the equivalent signifying the patient is at high risk of early progression/relapse of the prostate cancer indicating aggressiveness of the prostate cancer, or Low, Late or the equivalent, signifying that the patient is at low risk of early progression/relapse of the prostate cancer indicating indolence of the cancer.
A programmed computer operating as a classifier for predicting prostate cancer aggressiveness or indolence, comprising a processing unit and a memory storing a final classifier in the form of a set of feature values for a set of mass spectrometry features forming a constitutive set of mass spectra obtained from blood-based samples of prostate cancer patients, and a final classifier defined as a majority vote or average probability cutoff, of a multitude of master classifiers constructed from a combination of mini-classifiers with dropout regularization.
Dependencies are inferred from the claim text.