# US 12,009,059 B2 · Analytic prediction of antibiotic susceptibility

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

- Number: US 12,009,059 B2
- Document ID: US12009059B2
- Status: Granted
- Kind: B2
- Application number: 16/463,710
- Filed: 2017-11-27
- Published: 2019-09-12
- Granted: 2024-06-11
- Earliest claimed: 2016-11-28
- Assignees: Koninklijke Philips N.V.
- Applicants: Koninklijke Philips N.V.
- Inventors: Karthikeyan Murugesan
- Web page: https://patentgrep.com/patent/US12009059B2

## CPC

- G16B 20/00: ICT specially adapted for functional genomics or proteomics, e.g. genotype-phenotype associations
- G06N 20/00: Machine learning
- G16B 40/00: ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
- G16B 50/20: Heterogeneous data integration
- G16B 25/10: Gene or protein expression profiling; Expression-ratio estimation or normalisation
- G16B: BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Methods and systems for predicting the susceptibility of bacterial pathogens to antibiotics using genomic data sets. Various embodiments described herein receive a genomic dataset and a set of labels and run principal variance component analysis thereon to determine the effect sizes of the labels. One or more labels are then selected based on their effect sizes and used in a machine learning model to make predictions on future datasets.

## Claims

### Claim 1

A method of training a model for predicting antibiotic drug susceptibility, the method comprising:
receiving, via an interface, a data set of a plurality of features;
receiving, via the interface, a set of labels that are related to the plurality of features, wherein the set of labels is a covariate matrix including a plurality of phenotypes, and the plurality of phenotypes includes susceptibilities to various antibiotics and at least one of age, year, isolate collection date, isolate sequencing date, and variation in a sequence measured by multilocous sequence typing;
supplying the data set of the plurality of features and the set of labels to a processor configured to execute instructions stored on a memory to provide a variance analysis engine, wherein the variance analysis engine is configured to generate an effect size for each of the labels on the data set of the plurality of features;
supplying as output from the variance analysis engine to a sizing module at least one effect size generated for a label;
selecting, via the sizing module, at least one label to be used in a machine learning model based on the at least one supplied effect size; and
training, using the selected at least one label with a data set, a machine learning model to predict antibiotic drug susceptibility.

### Claim 2 (depends on 1)

The method of claim 1, wherein the data set of the plurality of features is a genomic dataset including at least one of a gene presence-absence matrix, an SNP matrix, a plasmid profiling matrix, a mobile genetic matrix, a gene expression matrix, an RNA sequence matrix, and a microarray matrix.

### Claim 3 (depends on 1)

The method of claim 1, wherein the set of labels is a single vector of binary values.

### Claim 4 (depends on 1)

The method of claim 1, wherein the set of labels is a single vector of multi-class values.

### Claim 5 (depends on 1)

The method of claim 1, wherein selecting the at least one label via the sizing module includes selecting the at least one label based on its generated effect size exceeding a predetermined threshold.

### Claim 6 (depends on 1)

The method of claim 1, further comprising ranking, via the sizing module, the plurality of labels based on their effect size and selecting, via the sizing module, the at least one label based on the ranking.

### Claim 7 (depends on 1)

The method of claim 1, wherein the data set of a plurality of features is a feature matrix including a plurality of genomic features and a plurality of isolates, and wherein the covariate matrix includes the plurality of phenotypes and the plurality of isolates.

### Claim 8

A system for training a model for predicting antibiotic drug susceptibility, the system comprising:
an interface for receiving a data set of a plurality of features and a set of labels that are related to the plurality of features; wherein the set of labels is a covariate matrix including a plurality of phenotypes, and the plurality of phenotypes includes susceptibilities to various antibiotics and at least one of age, year, isolate collection date, isolate sequencing date, and variation in a sequence measured by multilocous sequence typing;
a memory; and
a processor configured to execute instructions stored on the memory to: (i) provide a variance analysis engine configured to receive the data set of the plurality of features and the set of labels; (ii) output an effect size for each of the labels, wherein at least one label is selected to be used in a machine learning model based on its effect size; and (iii) train, using the selected at least one label with a data set, a machine learning model to predict antibiotic drug susceptibility.

### Claim 9 (depends on 8)

The system of claim 8, wherein the data set of the plurality of features is a genomic dataset including at least one of a gene presence-absence matrix, an SNP matrix, a plasmid profiling matrix, a mobile genetic element matrix, a gene expression matrix, an RNA sequence matrix, and a microarray matrix.

### Claim 10 (depends on 8)

The system of claim 8, wherein the set of labels is a single vector of binary values.

### Claim 11 (depends on 8)

The system of claim 8, wherein the set of labels is a single vector of multi-class values.

### Claim 12 (depends on 8)

The system of claim 8, wherein the at least one selected label is selected based on its generated effect size exceeding a predetermined threshold.

### Claim 13 (depends on 8)

The system of claim 8, wherein the variance analysis engine is further configured to rank the set of labels based on their effect size and select the at least one label based on the ranking.

### Claim 14 (depends on 8)

The system of claim 8, wherein the data set of a plurality of features is a feature matrix including a plurality of genomic features and a plurality of isolates, and wherein the covariate matrix includes the plurality of phenotypes and the plurality of isolates.