# US 2020/0350078 A1 · Library screening for cancer probability

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2020/0350078 A1
- Document ID: US20200350078A1
- Status: Published application
- Kind: A1
- Application number: 16/877,139
- Filed: 2020-05-18
- Published: 2020-11-05
- Earliest claimed: 2018-01-08
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Avijit Chatterjee; Wendy Wang
- Web page: https://patentgrep.com/patent/US20200350078A1

## CPC

- G16H 50/30: for calculating health indices; for individual health risk assessment
- G06N 5/00: Computing arrangements using knowledge-based models
- G16H 10/60: for patient-specific data, e.g. for electronic patient records
- G16H 15/00: ICT specially adapted for medical reports, e.g. generation or transmission thereof
- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G16H 50/50: for simulation or modelling of medical disorders
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method, system, and computer program product are provided for generating a predictive model. A processor(s) obtains a raw data set (peptide libraries) of patients designated as diagnosed/pre-diagnosed with a condition or not diagnosed with the condition. The processor(s) segments the raw data set into a pre-defined number of groups and separates out a holdout group. The processor(s) performs a principal component analysis on the remaining groups to identify, based on a frequency of features in the remaining groups, common features (principal components) in the remaining groups and weighs the common features based on frequency of occurrence. The processor(s) determines a smallest number of the principal components that yields a pre-defined level of validation accuracy. The processor(s) generates a predictive model, by utilizing the smallest number for a best fit in a logistic regression model. The predictive model provides binary outcomes.

## Claims

### Claim 1

A computer program product comprising:
a computer readable storage medium readable by one or more processors and storing instructions for execution by the one or more processor for performing a method comprising:
obtaining, by the one or more processors, a raw data set comprising peptide libraries of patients designated as either diagnosed or pre-diagnosed with a condition or not diagnosed with the condition;
segmenting, by the one or more processors, the raw data set into a pre-defined number of groups, wherein the segmenting comprises separating out a holdout group of data from the remaining groups;
performing, by the one or more processors, a principal component analysis on the remaining groups to identify, based on a frequency of features in the remaining groups, by the one or more processors, common features in data of the remaining groups and weighting the common features based on frequency of occurrence in the remaining groups, wherein the common features comprise principal components represented by coefficients;
determining, by the one or more processors, a smallest number of the principal components that yields a pre-defined level of validation accuracy;
generating, by the one or more processors, a predictive model, by utilizing the smallest number of principal components as parameters for a best fit in a logistic regression model, wherein the predictive model provides binary outcomes selected from the group consisting of: likelihood of presence of the condition within a pre-defined threshold or no likelihood of the presence of the condition within the pre-defined threshold;
obtaining, by the one or more processors, new raw data comprising a 12-mer peptide library of a given individual;
applying, by the one or more processors, the predictive model to the new raw data to determine, a binary outcome for the given individual; and
providing, by the one or more processors, the binary outcome to the given individual, in real-time.

### Claim 2 (depends on 1)

The computer program product method of claim 1, further comprising:
testing, by the one or more processors, the predictive model, by applying the coefficients to calculate features to use for the holdout group; and
tuning, by the one or more processors, the predictive model by comparing the features for the holdout group to the principal components.

### Claim 3 (depends on 1)

The computer program product method of claim 1, wherein the peptide libraries comprise 12-mer peptide libraries.

### Claim 4 (depends on 1)

The computer program product method of claim 1, wherein the condition comprises Stage I breast cancer.

### Claim 5 (depends on 1)

The computer program product method of claim 1, wherein the segmenting comprises generating random numbers to select which record of the raw data set to assign to which of the groups.

### Claim 6 (depends on 1)

The computer program product method of claim 1, comprising:
prior to performing the principal component analysis, normalizing, by the one or more processors, the peptides comprising the raw data set by adding peptide values of the peptides across all the peptides and dividing each resultant expression value of the peptide values of the peptides added across all the peptides by a resultant sum to compute a ratio.

### Claim 7 (depends on 1)

The computer program product method of claim 1, wherein performing a principal component analysis on the remaining groups comprises:
generating, by the one or more processors, sets of training data, wherein each set of training data of the sets of training data comprises data in all but one group of the remaining groups;
performing, by the one or more processors, a principal component analysis individually on each set of training data to identify common features in each set of training data;
executing, by the one or more programs, a principal component analysis utilizing the remaining groups as a single training set to identify common features in the single training set; and
performing a cross-validation of the common features of each set of training data and the common features in the single training set, wherein the cross-validation is of a number of folds equal to a number of the remaining groups, to identify the principal components and determine the coefficients.

### Claim 8 (depends on 7)

The computer program product method of claim 7, wherein the pre-defined level of validation accuracy is a highest validation accuracy based on the number of folds.

### Claim 9 (depends on 1)

The computer program product method of claim 1, wherein applying the predictive model comprises performing a principal component analysis on the new raw data utilizing the coefficients and performing a logistic regression to generate the binary outcome.

### Claim 10

A system comprising:
a memory;
one or more processors in communication with the memory; and
program instructions executable by the one or more processors via the memory to perform a method, the method comprising:
obtaining, by the one or more processors, a raw data set comprising peptide libraries of patients designated as either diagnosed or pre-diagnosed with a condition or not diagnosed with the condition;
segmenting, by the one or more processors, the raw data set into a pre-defined number of groups, wherein the segmenting comprises separating out a holdout group of data from the remaining groups;
performing, by the one or more processors, a principal component analysis on the remaining groups to identify, based on a frequency of features in the remaining groups, by the one or more processors, common features in data of the remaining groups and weighting the common features based on frequency of occurrence in the remaining groups, wherein the common features comprise principal components represented by coefficients;
determining, by the one or more programs, a smallest number of the principal components that yields a pre-defined level of validation accuracy; and
generating, by the one or more processors, a predictive model, by utilizing the smallest number of principal components as parameters for a best fit in a logistic regression model, wherein the predictive model provides binary outcomes selected from the group consisting of: likelihood of presence of the condition within a pre-defined threshold or no likelihood of the presence of the condition within the pre-defined threshold;
obtaining, by the one or more processors, new raw data comprising a 12-mer peptide library of a given individual;
applying, by the one or more processors, the predictive model to the new raw data to determine, a binary outcome for the given individual; and
providing, by the one or more processors, the binary outcome to the given individual, in real-time.

### Claim 11 (depends on 10)

The system of claim 10, the method further comprising:
testing, by the one or more processors, the predictive model, by applying the coefficients to calculate features to use for the holdout group; and
tuning, by the one or more processors, the predictive model by comparing the features for the holdout group to the principal components.

### Claim 12 (depends on 10)

The system of claim 10, wherein the peptide libraries comprise 12-mer peptide libraries.

### Claim 13 (depends on 10)

The system of claim 10, wherein the condition comprises Stage I breast cancer.

### Claim 14 (depends on 10)

The c system of claim 10, wherein the segmenting comprises generating random numbers to select which record of the raw data set to assign to which of the groups.

### Claim 15 (depends on 10)

The c system of claim 10, the method further comprising:
prior to performing the principal component analysis, normalizing, by the one or more processors, the peptides comprising the raw data set by adding peptide values of the peptides across all the peptides and dividing each resultant expression value of the peptide values of the peptides added across all the peptides by a resultant sum to compute a ratio.

### Claim 16 (depends on 10)

The system of claim 10, wherein performing a principal component analysis on the remaining groups comprises:
generating, by the one or more processors, sets of training data, wherein each set of training data of the sets of training data comprises data in all but one group of the remaining groups;
performing, by the one or more processors, a principal component analysis individually on each set of training data to identify common features in each set of training data;
executing, by the one or more programs, a principal component analysis utilizing the remaining groups as a single training set to identify common features in the single training set; and
performing a cross-validation of the common features of each set of training data and the common features in the single training set, wherein the cross-validation is of a number of folds equal to a number of the remaining groups, to identify the principal components and determine the coefficients.

### Claim 17 (depends on 16)

The system of claim 16, wherein the pre-defined level of validation accuracy is a highest validation accuracy based on the number of folds.

### Claim 18 (depends on 10)

The system of claim 10, wherein applying the predictive model comprises performing a principal component analysis on the new raw data utilizing the coefficients and performing a logistic regression to generate the binary outcome.
