Predictive test for melanoma patient benefit from interleukin-2 (IL2) therapy

1. A method for predicting whether a melanoma patient is likely to benefit from high dose IL2 therapy, comprising the steps of:

a) performing, by a mass spectrometer, mass spectrometry on a blood-based sample of the patient and obtaining mass spectrometry data of the sample;

b) performing, by a computer implementing a classifier, a classification of the mass spectrometry data obtained by the mass spectrometer, wherein the classifier is developed from a development set of samples from melanoma patients treated with the high dose IL2 therapy comprising:

iteratively training, by the computer, Classifier 1 from the development set of samples and a set of mass spectral features identified as being associated with an acute response biological function to generate an Early class label, a Late class label, or the equivalent for each of a subset of samples, and

iteratively training, by the computer, Classifier 2 from a subset of samples classified with the Late class label by Classifier 1 in the development set of samples to generate an Early class label, a Late class label or the equivalent,

c) supplying, from the computer to a programmed computer trained to predict whether the melanoma patient is likely to benefit from treatment with high dose IL2 therapy, the mass spectrometry data of the sample obtained in step a) to a classifier using Classifier 1 and Classifier 2, wherein:

the classifier represents a k-nearest neighbor (kNN) classification algorithm implemented by the programmed computer and is arranged as a hierarchical combination of (a) the Classifier 1 classifying the patient into either a first Late group or a first Early group or the equivalent, and (b) the Classifier 2 further classifying the first Late group into a second Late group or a second Early group, wherein the patient classified into the first Late group or the second Late group is predicted to be likely to benefit from high dose IL2 therapy,

d) determining a Late class label for the sample from Classifier 2; and

e) administering the high dose IL2 therapy responsive to determining the Late class label for the sample.

2. The method of claim 1 , wherein Classifier 1 and Classifier 2 use features for performing classification of the sample or a subset thereof as follows:

Classifier 1

Classifier 2

3085

3111

3244

3590

3444

3613

3842

3818

4590

3888

5158

4051

5177

4434

5570

4793

5720

5003

6589

5071

6809

5417

6890

5675

6995

5692

8478

5765

9208

5840

11913

5953

11938

6348

11965

7297

13781

8491

14543

8565

14595

9098

15509

10162

10339

10590

10734

10804

10838

11056

11443

11502

11531

11627

11686

11913

11938

12004

12290

12459

12738

12871

12963

13135

13179

13326

13366

13573

13642

13845

14302

14543

14787

18265

21173

21373

23049

24551.

3. A method of detecting class labels for a melanoma patient comprising performing the method of claim 1 , as well as classifying the sample of the patient with a classifier developed from mass spectral data of a set of blood based samples obtained from melanoma patients treated with an anti-PD-1 drug to generate a class label, wherein class labels are detected.

4. A method of detecting a class label for a melanoma patient on high dose IL2 therapy by performing, by a mass spectrometer, mass spectrometry on a blood based sample from the melanoma patient and obtaining, at a computer implementing a classifier, mass spectrometry data of the sample; performing a classification of the mass spectrometry data using the computer implementing the classifier, wherein the classifier is developed from a development set of blood based samples obtained from melanoma patients treated with an anti-PD-1 drug, comprising iteratively training the classifier from the development set of samples to generate a class label of Late or the equivalent and Early or the equivalent, supplying, from the computer implementing the classifier to a programmed computer trained to predict whether a melanoma patient is likely to benefit from treatment with the anti-PD-1 drug, the mass spectrometry data of the blood based sample from the melanoma patient to the classifier, wherein the patient having a class label of Late or the equivalent is predicted to be likely to benefit from the anti-PD-1 drug, determining the class label of Late or the equivalent for the sample using the classifier, and administering the anti-PD-1 drug responsive to determining the class label of Late of the equivalent for the sample.