WO2025147585A1 - Use of epigenetic and genomic biomarkers for assessing drug response to type 2 inflammatory diseases - Google Patents

Abstract

Identifying and refining clinically significant patient stratification is a critical step toward realizing the promise of precision medicine in a number of medical conditions such as asthma. In this context, we describe the construction unbiased patient stratification scores based on DNA methylation (DNAm) patterns and gene expression patterns and their utilization to refine the efficacy of hallmark biomarkers for predicting drug response in asthma patients, as well as patients having diseases/conditions caused by excessive Type 2 inflammation, such as atopic dermatitis.

Description

USE OF EPIGENETIC AND GENOMIC BIOMARKERS FOR ASSESSING DRUG RESPONSE TO TYPE 2 INFLAMMATORY DISEASES

CROSS REFERENCE TO RELATED APPLICATIONS

This application claims the benefit under 35 U.S.C. Section 119(e) of copending and commonly-assigned U.S. Provisional Patent Application No. 63/617,288, filed January 3. 2024 entitled “A DIAGNOSTIC EPIGENETIC SCORE FOR DETERMINING THE EFFECT OF BIOMARKERS FOR ASTHMA DRUG RESPONSE, the contents of which is incorporated by reference herein.

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH AND DEVELOPMENT

This invention was made with government support under R01HL155024, and R21HG013393, awarded by the National Institutes of Health. The government has certain rights in the invention.

TECHNICAL FIELD

Embodiments of the disclosure concern at least the fields of molecular biology and medicine.

BACKGROUND OF THE INVENTION

Type 2 inflammation is a systemic allergic response that involves the activation of immune cells. It affects a large proportion of people with asthma, especially those with uncontrolled or severe asthma. Asthma is the most common chronic condition in childhood¹ and is considered a heterogeneous disease with subtypes that respond differently to therapies². Identification of biomedically meaningful disease subtypes is, therefore, a critical step toward realizing the promise of precision medicine for asthma. Clinical heterogeneity may reflect variation in disease mechanisms across individuals due to genetics³, environmental exposures⁴, and the interaction between the two⁵. Multiple asthma endotypes with various underlying molecular mechanisms have been identified and linked to disease severity and clinical trajectory⁶’⁷ However, asthma treatment response remains heterogeneous, and disease control remains suboptimal for many patients^{6 8}.

Different asthma endotypes, such as type 2 (T2)-high and T2-low asthma, exhibit distinct pathophysiological mechanisms and inflammatory profiles, each associated with specific biomarkers that guide clinical decision-making⁹. For example, the T2-high asthma endotype, driven by type 2 effector cells (type 2 helper T, cytotoxic T. and innate lymphoid cells) that secrete interleukin (IL)-4, IL-5, and IL-13, is primarily characterized by eosinophilic inflammation, which is often accompanied by an elevated peripheral blood eosinophil count (BEC)¹⁰. In contrast, T2-low asthma lacks systemic eosinophilia and is, instead, characterized by neutrophilic or paucigranulocytic inflammation and is often linked to severe, treatment-resistant forms of the disease¹¹. Identifying a specific endotype by analyzing biomarkers such as BEC, sputum eosinophil percentage, fraction of exhaled nitric oxide (FeNO), and total serum immunoglobulin E (IgE) level allows clinicians to develop personalized treatment strategies¹².

The provision of insurance coverage for most of the advanced, yet costly, targeted biologic therapies — approximately $40,000 annually¹³ — relies on the presence of specific biomarkers associated with T2-high asthma¹⁴. However, clinical outcomes of biologic treatment vary among patients, and these biomarkers appear clinically imprecise in predicting treatment outcomes in some patients. For instance, while biologic therapy can achieve clinical remission, a study of severe asthma in adult patients reported remission in less than a quarter of the cases¹⁵; and in pediatric patients with elevated BEC and IgE, treatment with the IL-4 receptor antagonist Dupilumab only halves the risk of exacerbations compared to a placebo over a year², suggesting that its efficacy may be limited in certain subgroups of the targeted population.

Unrecognized clinical heterogeneity within existing endotype definitions limits the precision of known biomarkers. Differences in clinical outcomes are evident across patient groups, including racially and ethnically diverse populations. For instance, African American and Puerto Rican children exhibit a reduced bronchodilator response (BDR) to short-acting p2-agonists (SABAs)¹⁶, the standard treatment for acute bronchospasm, even when combined with inhaled corticosteroids (ICS)¹⁷. Together with variability in biomarker distribution among diverse populations¹⁸, asthma clinical heterogeneity can compromise medical care and hinder health equity due to biases in drug eligibility¹⁹.

In view of the above, there is a need in the art for additional methods and materials useful for obtaining information on patient responses to therapeutic regimens that are directed to conditions characterized by type II inflammation, for example asthma.

SUMMARY OF INVENTION

Embodiments of the invention disclosed herein include novel methodologies that can provide a "genomic score” that is useful to, for example, guide clinicians in treating conditions involving type II inflammation disorders, including pediatric asthma. Using observations of peripheral blood DNA methylation patterns in children with asthma, we found that only a subset of patients benefit from observing eosinophil counts and IgE levels as biomarkers for predicting which patients are likely to respond well or poorly to albuterol treatment. In other words, eosinophil counts and IgE indicate response to albuterol for only a subset of pediatric patients with asthma. Our personalized genomic score disclosed herein, calculated based on DNA methylation data assayed using a commercially available Illumina array, can identify whether eosinophil counts and IgE predict drug response for a given patient. Embodiments of the invention disclosed herein have been discovered to be better than the current clinical standard (observations of eosinophil counts and IgE levels) for identifying children with asthma who will have a good or poor response to current asthma therapies.

The invention disclosed herein has a number of embodiments. Embodiments of the invention include, for example, methods of obtaining information on patient response to a therapeutic regimen comprising administration of a composition selected to treat a pathological condition characterized by type II inflammation. These methods typically comprise observing methylation of genomic DNA obtained from a patient at a plurality of selected CpG sites; and then obtaining information on patient response to the therapeutic regimen using the methylation patterns observed at the plurality of CpG sites. In certain embodiments of the invention, the patient is a pediatric asthma patient, and the therapeutic regimen comprises the administration of a bronchodilating agent such as albuterol. In other embodiments of the invention, the therapeutic regimen comprises the administration of a type II inflammation inhibitor such as omalizumab, mepolizumab, reslizumab, benralizumab, dupilumab, or tezepelumab.

Embodiments of the invention further include methods of observing methylation of genomic DNA; these methods comprise obtaining genomic DNA from a patient, wherein the patient is selected to have a pathological condition characterized by type II inflammation. Typically, these methods include combining DNA methylation microarray comprising a plurality of polynucleotides having CpG methylation sites whose methylation status is useful to predict patient response to a therapeutic regimen directed to a pathological condition characterized by type II inflammation; and then observing the fraction of presence or absence of methylation at CpG sites within the genomic DNA (across all cells in a sample).

Embodiments of the invention also include DNA microarrays comprising a plurality of polynucleotides having CpG methylation sites whose methylation status is useful to predict patient response to a therapeutic regimen used to treat a pathological condition characterized by type II inflammation. Typically, in these methods, at least 10 of the plurality of polynucleotides are hybridized to genomic sequence polynucleotides selected to be obtained from a patient identified as a candidate for a therapeutic regimen comprising the administration of albuterol, omalizumab, mepolizumab, reslizumab, benralizumab, dupilumab, or tezepelumab.

Other objects, features, and advantages of the present invention will become apparent to those skilled in the art from the following detailed description. It is to be understood, however, that the detailed description and specific examples, while indicating some embodiments of the present invention, are given by way of illustration and not limitation. Many changes and modifications within the scope of the present invention may be made without departing from the spirit thereof, and the invention includes all such modifications.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1. Figure 1 provides violin plots showing the correlation between peripheral blood eosinophil counts and serum IgE levels with the bronchodilator drug response (BDR) to albuterol, specifically, baseline biomarker levels as response biomarkers. These Violin plots show the distributions of (a) imputed BEC (%), (b) BEC (cells/micro L) and (c) log total serum IgE levels stratified by BDR response >= 12% (Yes, otherwise No). Low Wilcoxon rank-sum p-values and high odds ratio (with 95 % confidence intervals) indicate the biomarker is predictive of BDR for patients in the respective percentile range. Results are also stratified (columns) by study (GALA II, SAGE II) or combined (COMBINED).

Figure 2. Figure 2 provides further violin plots showing a correlation between peripheral blood eosinophil counts and serum IgE levels with the bronchodilator drug response (BDR) to albuterol.

Figure 3. Figure 3 provides graphed eosinophil (Eos) count data showing blood eosinophil proportions (BEC) as drug response biomarkers along the DNAm patient stratification spectrum. Grouped boxplots comparing the distributions of imputed blood eosinophil proportions (BEC%) across different percentile ranges of the DNAm score, stratified by BDR responders and non-responders. Low Wilcoxon rank-sum p-values and high odds ratio (with 95% confidence intervals) indicate the biomarker is predictive of BDR for patients in the respective percentile range. We present results using pairs of boxplots, where each pair demonstrates the biomarker's predictive effect on drug response among patients in one quartile of the methylation scores, grouping patients based on the similarity of their methylation scores. In each figure, we present the results in a discovery dataset used to train the models (top row; GALA dataset; a Latino cohort) and replicate the results on a test dataset (bottom row SAGE; an African American cohort). Figure 4. Figure 4 provides graphed IgE level data as drug response biomarkers along the DNAm patient stratification spectrum. Grouped boxplots comparing the distributions of IgE levels across different percentile ranges of the DNAm score, stratified by BDR responders and non-responders. Low Wilcoxon rank-sum p-values and high odds ratio (with 95% confidence intervals) indicate the biomarker is predictive of BDR for patients in the respective percentile range. We present results using pairs of boxplots, where each pair demonstrates the biomarker's predictive effect on drug response among patients in one quartile of the methylation scores, grouping patients based on the similarity of their methylation scores. In each figure, we present the results in a discovery dataset used to train the models (top row; GALA dataset; a Latino cohort) and replicate the results on a test dataset (bottom row SAGE; an African American cohort).

Figure 5. Figure 5 shows PACA DNAm scores stratify patients along a continuum corresponding to heterogeneity in the clinical presentation of asthma, (a) Radar plots illustrating distribution shifts of asthma phenotypes across different percentile ranges of the DNAm score. Each point on the radial scale represents an average phenotypic value for the respective quantile range, normalized on a 0 to 1 scale within each cohort, (b) Pearson's coefficients and 95% confidence intervals (Cis) for the correlation between the DNAm scores and clinical covariates, (c) Linear regression coefficients and 95% Cis for the DNAm score as the outcome and BDR as the variable of interest, adjusted for demographic and clinical variables. Results are stratified by study (GALA II, SAGE II) or combined (COMBINED).

Figure 6. Figure 6 shows BEC and IgE as albuterol drug response biomarkers along the DNAm patient stratification spectrum, (a) Grouped boxplots comparing the distributions of imputed blood eosinophil proportions (BEC%; left) and (log) IgE levels (right) across different percentile ranges of the DNAm score, stratified by BDR responders and non-responders. Low Wilcoxon rank-sum p-values and high odds ratio (with 95% confidence intervals) indicate the biomarker is predictive of BDR for patients in the respective percentile range, (b) The empirical joint distribution of BEC% and (log) IgE in patients with top and bottom decile DNAm scores, stratified by BDR responders and non-responders. (c) Distribution shifts in baseline presentation of clinical phenotypes known to be associated with response to anti-T2 asthma biologic drugs. Results are presented across both GALA II and SAGE II for better statistical power. Error bars in line plots indicate the standard error of the mean (SEM). Low Wilcoxon rank-sum p-values indicate differences in phenotype presentation between patient groups in the respective percentile range of the DNAm score. EXBR: Exacerbation score in the last 12 months (0-6); Allergen: Number of positive reactions on a skin prick allergen test; OCS: Time since the last oral steroid prescription (a lower value indicates a more recent prescription).

Figure 7. Figure 7 shows that Type 2 high patients with high DNAm scores present several clinical variables that are known from multiple studies in the literature to associate with drug response. The left panel shows the clinical relevance and importance of T2 high heterogeneity. The right panel show s current T2 biologies on the market and their eligibility criteria (also relevant for insurance coverage) and the reason BEC and IgE are relevant in the clinic. Type 2 high patients with low DNAm scores have a clinical presentation that is negatively associated with drug response. Figures 10 show results with gene expression. These results show' that our DNAm score is correlated (in a w ay that cannot be explained by other variables, which w e accounted for in our analysis) with changes in gene expression that are known to be part of the Type 2 inflammatory pathway.

Figure 8. Figure 8shows that DNAm score captures know n and novel genes in T2h. The left panel shows the overlap of over 100 genes, that replicate in 2 different ethnically and ancestrally diverse cohorts and pass multiple testing corrections, with known T2 high pathways and our DNAm score. Green: genes that overlap, Purple: novel genes that are specifically uncovered by our DNAm score heterogeneity, Orange: genes that are specific to known T2 high endotype. The right panel provides a pictorial illustration of the statistical test performed to find overlapping a new' genes in the plot on the right. Figure 9. Figure 9 highlights genes which are currently existing targets for the drugs labeled in the white boxes in T2 high asthma. This provides additional evidence on the biology related to T2 high asthma endotype.

DETAILED DESCRIPTION OF THE INVENTION

In the description of embodiments, reference may be made to the accompanying figures which form a part hereof, and in which is shown by way of illustration a specific embodiment in which the invention may be practiced. It is to be understood that other embodiments may be utilized, and structural changes may be made without departing from the scope of the present invention. Many of the techniques and procedures described or referenced herein are well understood and commonly employed by those skilled in the art. Unless otherwise defined, all terms of art, notations and other scientific terms or terminology used herein are intended to have the meanings commonly understood by those of skill in the art to which this invention pertains. In some cases, terms with commonly understood meanings are defined herein for clarity and/or for ready reference, and the inclusion of such definitions herein should not necessarily be construed to represent a substantial difference over what is generally understood in the art.

All publications mentioned herein are incorporated herein by reference to disclose and describe aspects, methods and/or materials in connection with the cited publications. Certain aspects of the invention are discussed in Gorla et al., ‘EPIGENETIC PATIENT STRATIFICATION VIA CONTRASTIVE MACHINE LEARNING REFINES HALLMARK BIOMARKERS IN MINORITIZED CHILDREN WITH ASTHMA”; Res Sq [Preprint], 2024 Sep 13:rs.3.rs-5066762. doi: 10.21203/rs.3.rs-5066762/vl (hereinafter “Gorla et al.”).

Type 2 inflammation is a specific immune response characterized by the activation of certain immune cells (like eosinophils, mast cells, and T-helper 2 cells) that release inflammatory mediators. This response is involved in conditions like asthma, allergic rhinitis, and eczema. Eosinophil counts and IgE levels are standard and easy-to-collect blood biomarkers routinely used to inform clinicians in deciding treatment strategies for asthma patients. Insurance companies currently approve a new generation of biologic asthma drugs ($30-47k per year) based on eosinophil counts, IgE levels, and physicians' subjective assessments of asthma severity.

Identifying and refining clinically significant patient stratification is a critical step toward realizing the promise of precision medicine in asthma. As noted above, several peripheral blood hallmarks, including total peripheral blood eosinophil count (BEC) and immunoglobulin E (IgE) levels, are routinely used in asthma clinical practice for endotype classification and predicting response to state-of-the-art targeted biologic drugs. However, these biomarkers appear ineffective in predicting treatment outcomes in some patients, and they differ in distribution between racially and ethnically diverse populations, potentially compromising medical care and hindering health equity due to biases in drug eligibility.

As disclosed herein, we confirmed a known correlation between peripheral blood eosinophil counts and serum IgE levels with the bronchodilator drug response (BDR) to albuterol, the world's most prevalent and first-line asthma drug. Data illustrating this is shown in Figures 1 and 2. We have discovered that BEC and IgE are predictive of BDR to albuterol in only a subset of the patients, and a "DNAm score” can be used to identify those patients. Data illustrating this discovery is shown in Figures 3 and 4. We have discovered that our DNAm score can also be used to identify patients for which BEC and IgE are predictive of biologic drugs targeting Type 2 inflammation. Elevated BEC and IgE levels (identified via a simple blood test) in asthma patients are biomarkers of Type 2 inflammatory asthma, a form of asthma in which excessive Type 2 inflammation leads to disease pathogenesis in asthma. They, therefore, serve as biomarkers for recommending a biologic treatment for shutting down Type 2 inflammation. Among the patients with elevated BEC and IgE levels (i.e.. Type 2 high patients), the patients with high DNAm scores are the patients who are likely to respond to these drugs. The drugs targeting Type 2 inflammation are more interesting since - unlike albuterol - they are very expensive. Embodiments of the invention include the construction of unbiased patient stratification scores based on DNA methylation (DNAm) patterns and its utilization to refine the efficacy of hallmark biomarkers for predicting drug response in asthma patients, as well as patients having diseases/conditions caused by excessive Type 2 inflammation, such as atopic dermatitis (eczema). Embodiments of the invention disclosed herein include novel methodologies that observe CpG methylation patterns at specific sites, information that can then be used, for example, to provide a “genomic score⁷' that can be used to. for example, guide clinicians in treating certain conditions including pediatric asthma (e.g., in patients having a CpG methylation profile indicative of a poor drug response).

Embodiments of the invention include, for example, methods of obtaining information on patient response to a therapeutic regimen comprising administration of a composition selected to treat a pathological condition characterized by type II inflammation. These methods typically comprise observing methylation of DNA obtained from a patient at a plurality of selected CpG sites; and then obtaining information on patient response to the therapeutic regimen using the methylation observed at the plurality of CpG sites. In certain embodiments of the invention, the patient is an asthma patient such as a pediatric asthma patient, and the therapeutic regimen comprises the administration of a bronchodilating agent such as albuterol. Optionally, the patient is selected to have some non-European ancestry. In typical embodiments of the invention, methylation patterns are observed in order to obtain information that is indicative of a poor response to albuterol. In typical embodiments of the invention methylation is observed in genomic DNA obtained from leukocyte cells obtained from the patient. In some embodiments of the invention, the therapeutic regimen comprises the administration of a type II inflammation inhibitor such as omalizumab, mepolizumab, reslizumab. benralizumab, dupilumab or tezepelumab. In certain embodiments of the invention, observing includes applying a weighted average of methylation markers within the plurality of CpG sites. In some embodiments, observing CpG site methylation comprises an algorithm that leverages a Canonical Correlation Analysis (CCA) to observed patterns in a given target data subset (for example, asthma patients) that are not observable in background data (for example, healthy, non-asthma patients).

Embodiments of the invention also include DNA methylation microarrays comprising a plurality of polynucleotides having CpG methylation sites whose methylation status is useful to predict patient response to a therapeutic regimen used to treat a pathological condition characterized by type II inflammation. Typically, in these methods, at least 10 of the plurality of polynucleotides are hybridized to genomic sequence polynucleotides obtained from a patient identified as a candidate for a therapeutic regimen comprising the administration of albuterol, omalizumab, mepolizumab, reslizumab, benralizumab, dupilumab or tezepelumab. Optionally, the patient is a pediatric asthma patient. In certain embodiments of the invention, genomic sequence polynucleotides are found in segments (e.g., about 30, 40, 50 or 60 nucleotide segments) of genomic DNA obtained from leukocytes or epithelial cells obtained from the patient.

Embodiments of the invention further include methods of observing methylation of CpG sites in genomic DNA, these methods comprising obtaining genomic DNA from a patient, wherein the patient is selected to have a pathological condition characterized by type II inflammation. Typically these methods include combining the genomic DNA with a DNA microarray comprising a plurality of polynucleotides having CpG methylation sites whose methylation status is useful to predict patient response to a therapeutic regimen directed to a pathological condition characterized by type II inflammation; and then observing the presence or absence of methylation at CpG sites within the genomic DNA, wherein said observing comprises performing a bisulfite conversion process on the genomic DNA so that cytosine residues in the genomic DNA are transformed to uracil, while 5-methylcytosine residues in the genomic DNA are not transformed to uracil.

In illustrative working embodiments of the invention, the CpGs methylation sites comprise: cg09319072, cg25532677, cg!7059181. cg24718015, cg04084598, cgl0897651, cg22158248, cg24005645, cgl9854901, cg24652919, cg22221575, cg!9004267, cg!5452204, cg05156120c, g26666090, cgl8399629, cgO9558738, cg20805961, cg20263733, and/or cgl4617280. In some embodiments of the invention, the CpGs methylation sites can comprise the Eos-specific CpGs: cgl7883142 (CLIP1), cgl 8254848 (CLC), cg26724455 (VTI1 A), cg01901579 (DTCER1 ), cgl l 841710 (SYNDIG1), cg08348441 (GLYR1), cg27508506 (MAPK13), cg!4617280 (CREG1), and/or cg20263733 (ATP2C1). We tested our epigenetics score’s top 100 most informative CpGs for cell -type-1 ev el differential DNA methylation with our DNAm scores while accounting for demographics, technical variation, and cell-type composition. We identified the above 9 CpGs associated with the DNAm score in eosinophil cells. These CpGs, identified in GALA II (Bonferroni adjusted P<0.05) and replicated in SAGE II (P<0.012), present hypermethylation in eosinophil cells of samples with high DNAm scores. Note that the gene names in the parentheses above indicate the gene body or region these CpGs are located in. In this context, hypermethylation" means that the contribution of all these methylation sites should be summed. In other words, increased methylation levels in all of these sites is correlated with high levels of our DNAm scores.

In certain embodiments of the invention that use algorithms for observations of peripheral blood DNA methylation patterns in children with asthma, we discovered that only a subset of patients benefit from observing eosinophil counts and IgE levels as biomarkers for predicting which patients are likely to respond well or poorly to albuterol treatment. In other words, eosinophil counts and IgE indicate response to albuterol for only a subset of pediatric patients with asthma. In embodiments of the personalized genomic score methodologies disclosed herein, we include DNA methylation data assays using a commercially available Illumina array, to identify whether eosinophil counts and IgE predict drug response for a given patient. In doing so, we developed novel methodologies that have correlative/predictive power that is better than that observed with the current clinical gold standards (eosinophil counts and IgE levels) that are conventionally used for identifying children with asthma who will have a good (or poor response) to an asthma therapy.

In certain embodiments of the invention, we employed a contrastive machine learning algorithm for learning variation that exists in a specific group of interest (e.g., patients with a disease) but does not exist in a background group (e.g., healthy individuals without the disease (controls)). We applied our method for identifying biomedically-relevant sub-phenotypic variation (disease subtypes). We described the contrastive learning algorithm in Gorla et al.. ' PHENOTYPIC SUBTYPING VIA CONTRASTIVE LEARNING” bioRxiv [Preprint], 2023 Jan 6:2023.01.05.522921. doi: 10.1101/2023.01.05.522921. Building upon this disclosure, we subsequently discovered a number of surprising ways that this algorithm can be successfully applied in clinical contexts, including adapting/applying the algorithm to DNA methylation data from a pediatric study of participants with asthma.

One algorithm embodiment was used to develop a genomic (methylation) score for asthma patients based on DNA methylation data. This score calculates a linear combination of several thousands of methylation positions in the genome. Our novel genomic score can predict whether children with asthma will have a good or poor bronchodilator drug response to albuterol. For example. Figure 3 (showing eosinophil count data) and Figures 4 (showing IgE data) show that our score stratifies patients: these biomarkers are predictive of bronchodilator drug response (albuterol) only for patients with a low score and not for patients with a high score. We present results using pairs of boxplots, where each pair demonstrates the biomarker's predictive effect on drug response among patients in one quartile of the methylation scores, grouping patients based on the similarity of their methylation scores. In each figure, we present the results in a discovery dataset used to train the models (top row; GALA dataset; a Latino cohort) and replicate the results on a test dataset (SAGE; an African American cohort).

We demonstrate that our novel genomic score is more accurate at predicting high or low bronchodilator drug response to albuterol among children with asthma than conventional methods. In embodiments of the invention, a patient's genomic (methylation) score can be used to guide clinicians in determining which patients will respond well or poorly to albuterol treatment (i.e., selecting use, dosage, and frequency). In addition, the stratification based on our genomic score across the four quartiles of patients may be informative for treatment decisions beyond albuterol prescription. Our novel methodology based on methylation profiles further stratifies subgroups of children with asthma, who are likely to respond well or poorly to albuterol treatment and is an improvement on the cunent clinical diagnostic standard. Solely relying on eosinophil counts and IgE levels for drug eligibility, in that case, would misclassify patients with asthma and may lead to worse asthma outcomes, increased healthcare utilization, and increased financial costs. Our novel genomic score is more accurate than the current clinical standard at predicting children with asthma who are likely to respond well or poorly to albuterol treatment. Further aspects and embodiments of the invention are discussed below.

As disclosed herein, embodiments of the invention can be used to identify patients who are more likely to respond to type 2 therapies. In this context, insurance companies (and clinicians) can use embodiments of the invention to more accurately predict responders to expensive type 2 therapies. In addition, pharma companies can use embodiments of the invention to profile the non-responders and develop new, more effective treatments. In particular, embodiments of the invention don't only identify likely responders to type 2 therapies but also likely non-responders among the patient group considered to be Type 2 patients. This means that future drugs for the current non-responsive patient subgroup may utilize embodiments of the invention to determine who should get the new drugs. For example, embodiments of the invention can be used to identify a new subset of asthma patients who are considered Type 2 asthma patients (or "Type 2 high" asthma patients), that is, patients in which type 2 inflammation is the cause of their asthma. Type 2 asthma patients are identified as patients with high levels of eosinophil cells (BEC) in their blood (above 4% of the total composition of white blood cells) or patients with high IgE levels (total IgE above 100). Embodiments of the invention include observations of the expression of certain genes. For example, provided below are three lists of genes including: 1) Shared Genes, which are replicating (passing 0.05 FDR in both GALA and SAGE separately) genes that are shared between the canonical T2 endotype and PACA defined variation within the T2-high cases only; as well as 2) T2endo specific genes, which are genes which are specific to the canonical T2 endotype and replicate; and 3) T2hPACA specific Genes, which are genes which are specific to the PACA defined variation within the T2-high cases only and replicate. The shared genes encompass a lot of the known T2 genes and is a good positive indication that we're observing significantly meaningful biological variation. In addition, IL 17 -receptor, D0K2 and CD244 are specific to canonical endotype definition. Finally, the T2 high PACA specific variation set identifies genes that are associated with B-cell and possibly plasma cells.

GENE LISTS

A. Shared Genes

TNFRSF14-AS1 MTFR1L MARCKSL1 INPP5B ACOT11 CSF1 TMIGD3 ADORA3 OXER1 EPAS1 IL1RL1 INPPI SLC16A14 IL5RA SRGAP3 CAMK1 CCR3 HYAL3 HES1 TEC AC 111000.4 GAPT THBS4 RHOBTB3 RAB44 SLC29A1 MYB RPS6KA2 FBP1 GFI1B PYROXD2 PLEKHA7 ABTB2 CAT PTGDR2 ASRGL1 HRASLS5 LGALS12 P2RY2 AP002761.4 ACACB HRK CYSLTR2 CEBPE LINC01303 TTC7B ASB2 AL590327.1 SORD SEMA7A ADAMTS7P1 ADAMTS7P4 PRSS33 ACSM3 EEF2K HSD3B7 ADGRG5 SMPD3 AC099521.2 IL34 ATP2A3 SPNS3 ALOX15 MFSD6L PIK3R6 PMP22 EPN2 CCL23 ACSF2 AC025048.4 HRH4 ADGRE1 ADGRE4P PNPLA6 CYP4F12 CLC GIPR SIGLEC8 CACNG8 CACNG6 VSTM1 OLIG2 OLIG1 BACE2 TFF3 GATA1 RHOXF1P3

B. T2endo specific Genes AJAP1 PAFAH2 PIK3R3 GSTM4 P0LR3GL CD244 ID2 GRHL1 IL17RB SYNE1 D0K2 LINC01504 FGFR2 GALNT6 TINF2 DHRS1 PHKG2 SNX20 FAM157C LINC02073 AC144831.1 0LFM2 LAIR1 AJ239328.1 RPGR TKTL1

C. T2hPACA specific Genes

CEP85 ITM2C ALAS1 TXNDC5 AMPD3 IGHV6-1 ZNF597 TXNDC11 TNFRSF17 TNFRSF13B SIGLEC10 IGLV5-45

One list above is based on the "T2hPACA specific" genes. An increase in total expression across all of these genes is a score that is expected to tag responders to drugs targeting type 2 inflammation. In embodiments, we can "normalize" the expression of each gene (e.g., as a weighted combination of the gene expression values rather than a simple sum; we do not have the weights at this point though like we have for the methylation score). Embodiments of the invention can use only the genes TNFRSF13B, TNFRSF17, SIGLEC10, IGHV6-1, IGLV5-45. One subset of patients we identify is a subset of the current patients considered Type 2 asthma patients. Meaning, every time we say something like "patients with high score are more likely to respond to a drug" what we really mean is that patients with BOTH high score AND biomarkers indicating they are type 2 patients (in our case, based on elevated BEC or IgE) are more likely to respond. In such embodiments of the invention, the elevated BEC and IgE are part of the score/method. In embodiments of the invention, the gene signatures can be thought of as simple sums across genes. In other words, for a given list of genes above, an increased level in each one of the genes in the list is associated with high DNAm score, which means that an approximation of our DNAm score can be calculated by summing the expression across all genes in the list. In the list above, all significant genes except for FAM157C and RPGR have positive association (i.e. increased expression).

Embodiments of the invention use observations of CpG methylation patterns in combination with other observations, for example observations of the expression of certain genes listed above and/or observations of conventional T2-high biomarkers (e.g., observations of BEC and IgE).

Below we describe constructing an unbiased patient stratification score based on DNA methylation (DNAm) and utilizing it to refine the efficacy of hallmark biomarkers for predicting drug response. We developed Phenotype Aware Component Analysis (PACA), a novel contrastive machine-learning method for learning combinations of DNAm sites reflecting biomedically meaningful patient stratifications. Leveraging whole-blood DNAm from Latino (discovery; n=l,016) and African American (replication; n=756) pediatric asthma case-control cohorts, we applied PACA to refine the prediction of bronchodilator response (BDR) to the short-acting p2-agonist albuterol, the most used drug to treat acute bronchospasm worldwide. While BEC and IgE correlate with BDR in the general patient population, our PACA-derived DNAm score renders these biomarkers predictive of drug response only in patients with high DNAm scores. BEC correlates with BDR in patients with upper-quartile DNAm scores (OR 1.12; 95% CI [1.04, 1.22]; P=7.9 e-4) but not in patients with lower-quartile scores (OR 1.05; 95% CI [0.95, 1.17]; P=0.21); and IgE correlates with BDR in above-median (OR for response 1.42; 95% CI [1.24, 1.63]; P=3.9e-7) but not in below-median patients (OR 1.05; 95% CI [0.92, 1.2]; P=0.57). These results hold within the commonly recognized type 2 (T2)-high asthma endotype but not in T2-low patients, suggesting that our DNAm score primarily represents an unknown variation of T2 asthma. Among T2-high patients with high DNAm scores, elevated BEC or IgE also corresponds to baseline clinical presentation that is known to benefit more from biologic treatment, including higher exacerbation scores, higher allergen sensitization, lower BMI, more recent oral corticosteroids prescription, and lower lung function. Our findings suggest that BEC and IgE, the traditional asthma biomarkers of T2-high asthma, are poor biomarkers for millions worldwide. Revisiting existing drug eligibility criteria relying on these biomarkers in asthma medical care may enhance precision and equity in treatment. Current data-driven disease subtyping efforts predominantly rely on data clustering and latent representation techniques, ranging from classical unsupervised methods like K-means²⁴ and principal component analysis²⁵ to deep-leaming approaches such as autoencoders²⁶. These methods, whether explicitly or implicitly, rely on data-driven similarities between samples, which can obscure the identification of meaningful subtypes, especially when those subtypes are reflected by subtle signals in the data. Because these techniques analyze all features in the data, without additional constraints, they tend to emphasize dominant variation - sources of variation that correlate with many features in the data. In genomic data, however, these dominant sources of variation typically represent unwanted factors, such as cell-type composition²⁷, batch effects²⁸, or ancestry²⁹, rather than the true heterogeneity of the disease.

We developed Phenotype Aware Component Analysis (PACA), a novel contrastive machine-learning method for defining biomedically meaningful disease subtypes and patient stratifications from high-dimensional data (Supplementary Fig. la in Gorla et al.). Unlike standard approaches, PACA does not simply cluster patients based on the dominant variation in the data. Instead, PACA isolates disease-specific heterogeneity by first removing sources of variations that are shared with healthy control individuals from the disease case data, and then applying standard dimensionality reduction techniques (Supplementary Fig. la in Gorla et al.). By accounting for sources of variation present in both cases and controls, we expect the top axes of variation in case data to reflect disease heterogeneity. Since these shared sources of variation are generally unknown, we take an unsupervised approach to identify latent features that are consistent across the two groups.

We confirmed that PACA is statistically calibrated and better powered to identify disease heterogeneity compared to other contrastive methods^30,31 when applied to synthetic and real-world data, including gene expression, DNAm, and genotype data (Supplementary Fig. Sib and S2-S5; Supplementary Note S2 in Gorla et al.). Unlike PACA, existing contrastive learning methods^30,31 require tuning a vague contrastive hyperparameter, which controls the level of variation among cases compared to controls. Such hyperparameters are likely to be misspecified when seeking to learn unknown disease heterogeneity (Supplementary Note S3 in Gorla et al.), often leading to power loss or the false tagging of arbitrary patterns as real disease heterogeneity (Supplementary Fig. S6 in Gorla et al ).

An epigenetic patient stratification score associates with drug response in pediatric asthma cohorts

We utilized PACA to develop an asthma patient stratification score using wholeblood DNAm from the Genes -environments & Admixture in Latino Americans II (GALA II), a pediatric asthma cohort primarily of Mexican and Puerto Rican ethnicities (n=618 cases, n=398 controls; Supplementary Table 1 in Gorla et al.). To reduce noise and narrow the feature space in the data, we first evaluated all methylation sites (>730,000 cytosine-phosphate-guanine probes, or CpGs) for statistical association with asthma (see Methods). An analysis with PACA considering only CpGs that demonstrated a nominally significant association with asthma (P<0.01) resulted in a patient stratification model based on a linear combination of 7.662 CpGs. To confirm the cross-population consistency of the resulting asthma stratification, we applied the DNAm score to a pediatric African American patient replication cohort with wholeblood DNAm from the Study of African Americans, Asthma, Genes, and Environments (SAGE II; n=429 cases; Supplementary Table 1 in Gorla et al ).

In both GALA II and SAGE II, the DNAm score stratified patients along a continuum corresponding to heterogeneity in lung function and clinical presentation of asthma (Fig. la,b). Specifically, high DNAm scores are associated (Pearson correlation; Bonferroni adjusted P<0.05) with lower baseline lung function as measured by forced vital capacity (FVC). peak expiratory flow rate (PEF). forced expiratory volume in one second (FEV 1), and forced expiratory flow (FEF). Other asthma phenotypes associated with high DNAm scores include higher BDR, elevated BEC and IgE, and higher exacerbation scores (Methods; Fig. lb; Supplementary Tables S2 and S3 in Gorla et al.).

The associations with BDR, BEC, and IgE remained significant (linear regression; GALA II P<2.7e-6, SAGE II P<2.9e-3) after adjusting for demographics (age, sex, ancestry, and ethnicity), body mass index (BMI), education level, medication intake, and inhaled (ICS) and oral (OCS) corticosteroids use (Fig. 1c; Supplementary Fig. S7 in Gorla et al.). Inspecting the top 20 most informative CpGs of our stratification model revealed heavily weighted CpGs in genes implicated in asthma pathogenesis and regulation of airway inflammation and remodeling (STATS. RASSFT ME )X I)'¹ '⁴. bronchodilator response (DZW5 )³⁵, and asthma severity and lung function (ALDH2) ⁶³¹ (Supplementary⁷ Tables S5 and S6 in Gorla et al.). We confirmed that existing contrastive learning methods did not yield similarly meaningful asthma patient stratification (Supplementary Table S2 and S4; Supplementary Note S2.7 in Gorla et al.).

BEC and IgE predict drug response only in patients with high DNAm scores

We evaluated whether our patient stratification model can explain the heterogeneity of biomarkers in predicting clinical outcomes, focusing on BDR to the short-acting p2-agonist albuterol, the most used drug for acute bronchospasm worldwide. The widely used asthma biomarkers BEC and IgE, both found to be associated with our DNAm score (Fig. 1), are known to be predictive of BDR³⁸. Across GALAII and SAGE II, the odds ratio (OR) for drug response (BDR>12%) is 1. 14 (95% CI [1.1, 1.18]) and 1.23 (95% CI [1.12, 1.35]) for BEC and IgE, respectively (Supplementary⁷ Fig. S8 in Gorla et al.).

We next investigated if the epigenetic score stratifies the population such that these biomarkers are predictive of BDR only in specific patient subgroups along the stratification spectrum (i.e., a statistical interaction). A linear regression model for BDR as the outcome revealed a significant interaction between IgE levels and the DNAm score (GALA II P=4.8e-3: SAGE II P=3.3e-6) after adjusting for IgE levels, demographics, BMI, education level, medication intake, ICS, and OCS (Supplementary Fig. S9 in Gorla et al.). Similarly, we found a significant interaction between BEC and the DNAm score in GALA II (P=5.2e-5; Supplementary Fig. S9 in Gorla et al.). This result was not replicated in SAGE II, likely owing to the limited number of BEC measurements available in this cohort (n=83). Even so, imputing missing BEC proportions from DNAm levels³⁹ (Supplementary' Fig. S10 in Gorla et al.) revealed a significant interaction between BEC and our DNAm score, which could not be explained by the BEC composition itself (GALA II P=8.3e-3, SAGE II P=4.4e-4; Supplementary Fig. S9 in Gorla et al.).

The statistical interactions suggest that elevated BEC and IgE predict BDR only in patients with high DNAm scores. In GALA II, classifying patients into responders (BDR>12%) and non-responders, IgE levels correlate with drug response in patients with above-median methylation scores (OR for response 1.42; 95% Cl [1.24, 1.63]; P=3.9e-7) but not in patients with below-median scores (OR 1.05; 95% CI [0.92, 1.2]; P=0.57); (imputed) BEC correlates with response in upper-quartile (OR 1.12; 95% CI [1.04. 1.22]; P=7.9 e-4) but not in lower-quartile patients (OR 1.05; 95% CI [0.95, 1.17]; P=0.21) (Fig. 2). More generally, we observe a linear increase in the correlation of BEC and IgE with response along the DNAm score continuum. These results were replicated in SAGE II (Fig. 2); results combining both datasets are provided in Supplementary Fig. Sil in Gorla et al.. Since low baseline lung function may explain BDR³⁸, we further verified that our results hold when restricting the analysis to patients with low baseline lung function (Supplementary Fig. S12 in Gorla et al.; Methods). Overall, our results confirm a robust monotonous increase in the effectiveness of elevated BEC and IgE as predictive drug response biomarkers along the patient stratification spectrum.

Next, we evaluated the performance of logistic regression models using both BEC and IgE as two-biomarker models for BDR prediction (Supplementary Table S7 in Gorla et al.). A standard two-biomarker model based on the entire patient population yielded ROC AUC 0.62 and 0.69 on GALA II (train) and SAGE II (test), respectively. In contrast, a two-biomarker model for patients with top-decile DNAm scores achieved ROC AUC of 0.76 and 0.75 in GALA II and SAGE II, respectively. As expected, a two- biomarker model for patients with bottom-decile DNAm scores, for which BEC and IgE are not associated with BDR, performed poorly (ROC AUC of 0.62 and 0.59). Evaluating other percentile ranges demonstrated a linear increase in performance along the DNAm score continuum (Supplementary' Table S7 in Gorla et al.).

We further evaluated alternative predictive models for BDR. A logistic regression model using cell-type composition, a significant source of variation in DNAm data⁴⁰, underperformed compared to the two-biomarker models tailored for patients with high DNAm scores (Supplementary Table S7 in Gorla et al.). A DNAm- based regularized logistic regression model using the same 7,662 CpGs that define our DNAm score outperformed the two-biomarker models in predicting BDR in GALA II (ROC AUC 0.83). However, unlike the simple two-biomarker model for patients with high DNAm scores, the performance of the DNAm-based predictive model did not replicate in the SAGE II cohort (ROC AUC 0.65). The likely overfitting to confounding effects, presumably leading to this poor consistency across cohorts, underscores PACA’s effectiveness in eliminating unknown confounding factors. This is achieved by using contrastive learning to remove sources of variation that are shared across cases and controls. Interestingly, both the cell-type composition and DNAm-based regularized logistic regression models performed better when evaluated in patient groups with high DNAm scores, further underscoring the robustness of the proposed DNAm score in stratifying clinical outcomes (Supplementary Table S7 in Gorla et al.).

The DNAm patient stratification score reflects heterogeneity of T2 asthma

The association of high DNAm scores with T2-high asthma (elevated BEC and IgE; Fig. l b) could suggest that our DNAm score simply recapitulates known T2 endotypes. However, the DNAm score stratified the effectiveness of BEC and IgE in predicting BDR even when restricting the analysis only to T2-high patients, as defined by BEC-high or IgE-high levels (Supplementary Fig. S12 in Gorla et al.; Methods). In contrast, this was not the case when considering only T2-low patients (Supplementary Fig. S13 in Gorla et al ), suggesting that our patient stratification primarily represents unknown variation that pertains to T2 asthma.

Most existing biologic treatments for asthma target the T2 pathway. Given this, we explored whether elevated BEC and IgE levels might be more suggestive of response to biologic drugs in patients with a high DNAm score, similar to their predictive value for BDR. We stratified patients based on their BEC and IgE levels, which are primary biomarkers of response to T2-targeted biologic drugs. We first observed that the DNAm score is associated with a difference in clinical presentation only among high-BEC or high-IgE patients, providing further evidence that our DNAm score reflects a spectrum of variation within the T2-high asthma endotype (Fig. 2c). Second, among patients with high DNAm scores, we observed that elevated BEC or IgE is associated with baseline clinical presentation that is known to benefit more from biologic treatment, including higher exacerbation scores⁴¹⁴², higher allergen sensitization⁴³, lower BMI⁴⁴⁴⁵, more recent OCS prescription⁴⁵’⁴⁶, and lower FEV1/FVC ratio⁴¹. In contrast, these characteristics are not as marked among patients with low DNAm scores, providing strong evidence that BEC and IgE are better predictors of biologic drug response among patients with high DNAm scores.

Finally, we asked whether epigenetic modifications specific to T2-high asthma drive our DNAm-based patient stratification and whether such modifications can be attributed to certain cell types. Using a deconvolution method for bulk DNAm⁴⁷, we tested our model’s top 20 most informative CpGs for cell-type-level differential DNAm with the DNAm scores while accounting for demographics, technical variation, and cell-type composition. We identified three CpGs associated with the DNAm score in eosinophil cells: cg!8399629, cg20263733, and cg!4617280 (Supplementary Table S8 in Gorla et al.). These CpGs, identified in GALA II (Bonferroni adjusted P<0.05) and replicated in SAGE II (P<0.012), present hypermethylation in eosinophil cells of samples with high DNAm scores; other cell types do not present differential methylation in these CpGs. Breaking the analysis into T2-high and T2-low patient groups (Methods) showed the statistical signal is driven by the T2-high patients (Supplementary- Table S9 and S10 in Gorla et al ), suggesting that eosinophils-specific differential epigenetic programming may explain the variation within T2 asthma identified by our DNAm stratification model.

Our findings indicate that DNAm scores have the potential to enhance the clinical utility- of established biomarkers and identify appropriate patient subgroups, leading to more personalized asthma management strategies. Among patients with low DNAm scores, the traditional asthma biomarkers BEC and IgE are ineffective for predicting BDR and may also be less useful for identifying those who would benefit from other therapies targeting T2-high asthma. Embracing a molecular approach based on epigenetics that integrates genetic and environmental factors, our DNAm-based asthma stratification score proves robust across populations and patient groups defined by established asthma phenotypes.

The DNAm score moves away from categorizing childhood asthma into discrete phenotypic or endotype clusters, instead highlighting that treatment response and biomarker effectiveness can exist along a predictable continuum. This approach aligns with the growing call for molecular profiles that define asthma phenotypes and endotypes more precisely, allowing for targeted "personalized" therapies⁴⁸. Our findings suggest that asthma heterogeneity, particularly within the T2-high endotype, is more complex than previously recognized.

Clinical responses to T2-targeting biologies for asthma vary; even among patients with similar eligibility criteria^2,15, highlighting the influence of multiple factors on asthma manifestation and treatment response. The apparent complexify of the T2- high endotype highlights the potential for identifying novel sub-phenotypes and developing more tailored treatment strategies. In particular, our results serve as a proof- of-principle for contrastive learning and epigenetics to robustly identify responders and possibly "super-responders"^42,45 to anti-T2 biological agents among children with suboptimal responses to therapy, regardless of their perceived endotype status. Integrating whole-blood epigenome-wide DNA methylation data with our novel contrastive machine learning algorithm enhances our understanding of the complex pathophysiological processes underlying asthma and help improve treatment strategies beyond response to bronchodilators. The association of reversible airway obstruction with eosinophilic inflammation³⁸, a hallmark of T2 asthma, suggests a shared pathophysiology, indicating that patients who respond to albuterol will also respond well to biologies. We observe that samples with high DNAm scores are enriched with eosinophil cells exhibiting hypermethylated sites near genes implicated in T2-high asthma and biologic drug response, including CREG1 and MIR4765^49,50. The conjectured epigenetically encoded memory in these eosinophil cells, as reflected by our DNAm score, may therefore align with the mechanisms targeted by other asthma therapeutics⁵¹. Furthermore, combining BEC and IgE with our DNAm score allowed us to identify a group of patients with a clinical profile known to respond to these drugs¹⁵-⁴¹-^{4244 46}.

References

1. Johnson, C. C. et al. Pediatric asthma incidence rates in the United States from 1980 to 2017. J. Allergy Clin. Immunol. 148, 1270-1280 (2021).

2. Bacharier, L. B. et al. Dupilumab in Children with Uncontrolled Moderate-to- Severe Asthma. N. Engl. J. Med. 385, 2230-2240 (2021).

3. Suzuki, K. et al. Genetic drivers of heterogeneity in type 2 diabetes pathophysiology. Nature 627, 347-357 (2024).

4. von Mutius, E. Environmental factors influencing the development and progression of pediatric asthma. J. Allergy Clin. Immunol. 109, S525-S532 (2002).

5. Hernandez-Pacheco, N., Kere, M. & Melen, E. Gene-environment interactions in childhood asthma revisited; expanding the interaction concept. Pediatr. Allergy Immunol. 33, el3780 (2022).

6. Wenzel, S. E. Asthma phenotypes: the evolution from clinical to molecular approaches. Nat. Med. 18, 716-725 (2012).

7. Woodruff, P. G. et al. T-helper Type 2-driven Inflammation Defines Major Subphenotypes of Asthma. Am. J. Resplr. Cril. Care Med. 180, 388-395 (2009).

8. Anderson, G. P. Endotyping asthma: new insights into key pathogenic mechanisms in a complex, heterogeneous disease. The Lancet 372, 1107-1119 (2008).

9. Hammad. H. & Lambrecht, B. N. The basic immunology of asthma. Cell 184, 2521-2522 (2021).

10. Ramakrishnan, R. K., Al Heialy, S. & Hamid, Q. Role of IL-17 in asthma pathogenesis and its implications for the clinic. Expert Rev. Resplr. Med. 13, 1057-1068 (2019).

11. Ray, A. & Kolls, J. K. Neutrophilic Inflammation in Asthma and Association with Disease Severity. Trends Immunol. 38, 942-954 (2017).

12. Licari, A., Manti, S., Castagnoli, R., Leonardi, S. & Marseglia, G. L. Measuring inflammation in paediatric severe asthma: biomarkers in clinical practice. Breathe 16, (2020).

13. Xu, X. et al. A cost comparison of benralizumab, mepolizumab, and dupilumab in patients with severe asthma: A US third-party payer perspective.

J. Manag. Care Spec. Pharm. 29, 1193-1204 (2023).

14. Tice, J. A. et al. Biologic Therapies for Treatment of Asthma Associated with Type 2 Inflammation: Effectiveness. Value, and Value-Based Price Benchmarks. https://icer.org/wp-content/uploads/2020/10/ICER_Asthma- Final-Report_Unredacted_08122020.pdf (2018).

15. Hansen, S. et al. Clinical Response and Remission in Patients With Severe Asthma Treated With Biologic Therapies. Chest 165, 253-266 (2024).

16. Burchard, E. G. et cd. Lower Bronchodilator Responsiveness in Puerto Rican than in Mexican Subjects with Asthma. Am. J. Respir. Crit. Care Med. 169, 386-392 (2004). 17. Samedy-Bates, L.-A. et al. Racial/Ethnic-Specific Differences in the Effects of Inhaled Corticosteroid Use on Bronchodilator Response in Patients With Asthma. Clin. Pharmacol. Ther. 106. 1133-1140 (2019).

18. Zhang, E., Levin. A. M. & Williams. L. K. How does race and ethnicity effect the precision treatment of asthma? Expert Rev. Precis. Med. Drug Dev. 4, 337- 356 (2019).

19. Burchard. E. G. & Borrell, L. N. Need for Racial and Ethnic Diversity in Asthma Precision Medicine. N. Engl. J. Med. 385. 2297-2298 (2021).

20. Tobias, D. K. et al. Second international consensus report on gaps and opportunities for the clinical translation of precision diabetes medicine. Nat. Med. 29, 2438-2457 (2023).

21. Vieira Braga, F. A. et al. A cellular census of human lungs identifies novel cell states in health and in asthma. Nat. Med. 25. 1153-1163 (2019).

22. Kachroo, P. et al. Metabolomic profiling reveals extensive adrenal suppression due to inhaled corticosteroid therapy in asthma. Nat. Med. 28, 814-822 (2022).

23. Han, R. et al. Decoding the role of DNA methylation in allergic diseases: from pathogenesis to therapy. Cell Biosci. 14, 89 (2024).

24. Ahlqvist, E. et al. Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables. Lancet Diabetes Endocrinol. 6, 361-369 (2018).

25. Zhu, H. et al. Identification of 6 dermatomyositis subgroups using principal component analysis-based cluster analysis. Int. J. Rheum. Dis. 22, 1383-1392 (2019).

26. Karim, M. R. el al. Deep learning-based clustering approaches for bioinformatics. Brief. Bioinform. 22. 393-415 (2021).

27. Rahmani, E. et al. Sparse PCA corrects for cell type heterogeneity' in epigenome-wide association studies. Nat. Methods 13, 443 (2016).

28. Barfield, R. T. et al. Accounting for population stratification in DNA methylation studies. Genet. Epidemiol. 38, 231-241 (2014). Rahmani, E. et al. Genome-wide methylation data mirror ancestry' information. Epigenetics Chromatin 10, 1-12 (2017). Abid. A., Zhang, M. J.. Bagaria, V. K. & Zou, J. Exploring patterns enriched in a dataset with contrastive principal component analysis. Nat. Commun. 9, 2134 (2018). Abid, A. & Zou, J. Contrastive variational autoencoder enhances salient features. ArXiv Prepr. ArXiv 190204601 (2019). Yang, X. O. et al. The signaling suppressor CIS controls proallergic T cell development and allergic airway' inflammation. Nat. Immunol. 14, 732-740 (2013). Sun, K. et al. Lnc-Clic5 as a sponge for miR-212-5p to inhibit cow bam PM2.5-induced apoptosis in rat alveolar macrophages. Toxicology 504, 153797 (2024). Yatagai, Y. et al. Expression quantitative trait loci for ETV4 and ME0X1 are associated with adult asthma in Japanese populations. Sci. Rep. 11, 18791 (2021). Joo, J. et al. Genome-wide association study in minority children with asthma implicates DNAH5 in bronchodilator responsiveness. Sci. Rep. 12, 12514 (2022). Au Yeung, S. L. et al. Aldehyde dehydrogenase 2-a potential genetic risk factor for lung function among southern Chinese: evidence from the Guangzhou Biobank Cohort Study. Awn. Epidemiol. 24, 606-611 (2014). Lee, H. W. et al. Peripheral blood transcriptomic clusters uncovered immune phenotypes of asthma. Respir. Res. 23, 237 (2022). Coverstone, A. M. et ai. Clinical significance of the bronchodilator response in children with severe asthma. Pediatr. Pulmonol. 54, 1694-1703 (2019). Rahmani, E. et al. BayesCCE: a Bayesian framework for estimating cell-type composition from DNA methylation without the need for methylation reference. Genome Biol. 19, 141 (2018).

40. Jaffe, A. E. & Irizarry R. A. Accounting for cellular heterogeneity⁷ is critical in epigenome-wide association studies. Genome Biol. 15, R31 (2014).

41. Chen. W. et al. Can we predict who will benefit most from biologies in severe asthma? A post-hoc analysis of two phase 3 trials. Respir. Res. 24, 120 (2023).

42. Portacci, A. et al. Real-world characteristics of “super-responders” to mepolizumab and benralizumab in severe eosinophilic asthma and eosinophilic granulomatosis with polyangiitis. ERJ Open Res. 9, (2023).

43. Sheehan, W. J. et al. Aeroallergen Sensitization, Serum IgE, and Eosinophilia as Predictors of Response to Omalizumab Therapy during the Fall Season among Children with Persistent Asthma. J. Allergy Clin. Immunol. Pract. 8, 3021-3028. e2 (2020).

44. Sposato, B., Bianchi. F., Ricci. A. & Scalese, M. Clinical Asthma Remission Obtained with Biologies in Real Life: Patients’ Prevalence and Characteristics. J. Pers. Med. 13, 1020 (2023).

45. Harvey, E. S. et al. Mepolizumab effectiveness and identification of superresponders in severe asthma. Eur. Respir. J. 55, (2020).

46. Bleecker, E. R. et al. Baseline patient factors impact on the clinical efficacy of benralizumab for severe asthma. Eur. Respir. J. 52, (2018).

47. Rahmani, E. et al. Cell-type-specific resolution epigenetics without the need for cell sorting or single-cell biology. Nat. Commun. 10, 3417 (2019).

48. Tyler, S. R. & Bunyavanich, S. Leveraging -omics for asthma endotyping. J. Allergy Clin. Immunol. 144, 13-23 (2019).

49. Hoang, T. T. et al. Epigenome-wide association study of DNA methylation and adult asthma in the Agricultural Lung Health Study. Eur. Respir. J. 56, 2000217 (2020).

50. Van Hulst, G. et al. Anti-IL5 mepolizumab minimally influences residual blood eosinophils in severe asthma. Eur. Respir. J. 59, 2100935 (2022).

51. Belanger, E. & Laprise, C. Could the Epigenetics of Eosinophils in Asthma and Allergy' Solve Parts of the Puzzle? Int. J. Mol. Set. 22, 8921 (2021).

All publications mentioned herein (e.g. those listed numerically herein) are incorporated herein by reference to disclose and describe the methods and/or materials in connection with which the publications are cited. Publications cited herein are cited for their disclosure prior to the filing date of the present application. Nothing here is to be construed as an admission that the inventors are not entitled to antedate the publications by virtue of an earlier priority date or prior date of invention. Further, the actual publication dates may be different from those shown and require independent verification. The following references include descriptions of methods and materials in this field of technology'.

CONCLUSION

This concludes the description of the illustrative embodiments of the present invention. The foregoing description of one or more embodiments of the invention has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many modifications and variations are possible in light of the above teaching.

Claims

1. A method of obtaining information on patient response to a therapeutic regimen comprising administration of a composition selected to treat a pathological condition characterized by type II inflammation, the method comprising: observing methylation of genomic DNA obtained from a patient at a plurality of CpG sites; and obtaining information on patient response to the therapeutic regimen using the methylation patterns observed at the plurality of CpG sites.

2. The method of claim 1, wherein the patient is diagnosed with asthma, and the therapeutic regimen comprises the administration of a bronchodilating agent.

3. The method of claim 2, wherein the bronchodilating agent is albuterol.

4. The method of claim 3, wherein the information is indicative of a poor response to albuterol.

5. The method of claim 2, wherein the asthma patient is a pediatric asthma patient.

6. The method of claim 2, wherein the patient is selected to have a non-European ancestry.

7. The method of claim 1, wherein methylation is observed in genomic DNA obtained from leukocytes obtained from the individual.

8. The method of claim 1, wherein observing includes applying a weighted average of methylation markers within the plurality of CpG sites.

9. The method of claim 1, wherein the therapeutic regimen comprises the administration of omalizumab, mepolizumab, reslizumab, benralizumab, dupilumab or tezepelumab.

10. A DNA microarray comprising a plurality of polynucleotides having CpG methylation sites whose methylation status is useful to predict patient response to a therapeutic regimen directed to a pathological condition characterized by type II inflammation, wherein at least 10 of the plurality of polynucleotides are hybridized to genomic sequence polynucleotides obtained from a patient identified as a candidate for a therapeutic regimen comprising the administration of albuterol, omalizumab, mepolizumab, reslizumab, benralizumab, dupilumab or tezepelumab.

11. The DNA microarray of claim 10. wherein the patient is a pediatric asthma patient.

12. The DNA microarray of claim 10, wherein genomic sequence polynucleotides are present in segment of genomic DNA obtained from leukocytes or epithelial cells obtained from the patient.

13. A method of observing methylation of genomic DNA comprising: obtaining genomic DNA from a patient, wherein the patient is selected to have a pathological condition characterized by type II inflammation; combining the genomic DNA with a DNA methylation microarray comprising a plurality' of polynucleotides having CpG methylation sites whose methylation status is useful to predict patient response to a therapeutic regimen directed to a pathological condition characterized by type II inflammation; and observing the presence or absence of methylation at CpG sites within the genomic DNA, wherein said observing comprises performing a bisulfite conversion process on the genomic DNA so that cytosine residues in the genomic DNA are transformed to uracil, while 5-methylcytosine residues in the genomic DNA are not transformed to uracil.

14. The method of claim 13. wherein the genomic DNA is obtained from a patient identified as a candidate for a therapeutic regimen comprising administration of albuterol, omalizumab, mepolizumab, reslizumab, benralizumab, dupilumab or tezepelumab.

15. The method of claim 13, wherein the patient is a pediatric asthma patient