Prediction of bumetanib treatment response in patients with autism spectrum disorder

Abstract

The present invention provides a method of predicting a response of a subject suffering from autism spectrum disorder (ASD) to bumetanib based on baseline expression levels of a panel of cytokines in the subject, baseline behavioral performance of the subject, and clinical information of the subject, and uses thereof. The invention further provides a prediction device for executing the prediction method and a computer readable medium.

Description

[0001] Related Applications

[0002] This invention claims the priority of the PCT application with an international filing date of April 14, 2023 and an international application number of PCT / CN2023 / 088307, which is incorporated herein in its entirety. Technical Field

[0003] This invention relates to the treatment of autism spectrum disorder (ASD), and specifically to biomarkers for predicting the response of ASD patients to bumetanide. Background Art

[0004] Autism spectrum disorder (ASD) affects approximately 1% of children globally 1 , which may lead to lifelong disabilities and increased premature mortality. Currently, there is no drug that can cure ASD or all of its core symptoms 2 . Recently, the successful cases of repurposing drugs for new psychiatric therapies have attracted people's attention 3 , and one example is the use of bumetanide to improve the main symptoms of ASD 4-8 . The most common adverse reactions are hypokalemia, increased urine output, loss of appetite, dehydration, and fatigue. There is significant heterogeneity in the efficacy of bumetanide in treating ASD patients, with heterogeneity ranging from 37.3% to 47.62% in domestic randomized controlled trials (RCTs) 5,8 , and 51.80% in French RCTs 6 or 26.3% - 45.2% 7 . In addition, some studies have reported that bumetanide has no significant effect on the treatment of ASD patients 9 . Understanding this heterogeneity is crucial for its clinical applicability, and further research on its potential mechanism of action is needed to achieve precision medicine for children with ASD.

[0005] The application of bumetanide as a potential drug for improving ASD symptoms is based on a hypothesized pathology of ASD, namely, the delayed developmental switch of γ-aminobutyric acid (GABA) from an excitatory function to an inhibitory one 10-12 . In valproic acid and fragile X rodent autism models, promoting GABA conversion can be achieved by reducing intracellular chloride concentration, and the reduction of intracellular chloride concentration is mediated by the continuous expression of major chloride transporters, such as potassium (K)-chloride cotransporter 2 (KCC2) and importin Na-K-Cl cotransporter 1 (NKCC1) 12 . Therefore, as an NKCC1 inhibitor, the ability of bumetanide to restore GABA function in ASD has been tested 5-7,13,14However, these transporters can also be affected by other molecules, such as cytokines, which are many small cell signaling proteins that interact closely to regulate the immune response. Cytokines are not only related to brain development 15 , but also related to GABAergic transmission 16-18 . It has been reported that interferon (IFN)-γ can reduce the levels of the α subunit of NKCC1 and Na+-K+-ATPase, contributing to the restoration of the function of inhibitory GABA 16 . In maternally deprived mice, interleukin (IL)-1 has also been found to reduce the expression of KCC2, delaying the developmental switch of GABA function, which may lead to the pathophysiology of developmental disorders such as ASD 17,18 . Therefore, a natural question is: whether the therapeutic effect of bumetanide on ASD will be affected by the patient's immune response.

[0006] In fact, compared with healthy controls, the cytokine levels in ASD patients have changed 19-22 . A recent meta-analysis showed that the levels of anti-inflammatory cytokines IL-10 and IL-1 receptor antagonist (Ra) in the blood of ASD patients were reduced 20 , while the pro-inflammatory cytokines IL-1β, IL-6 and the anti-inflammatory cytokines IL-4, IL-13 were elevated 21 . Elevated levels of IFN-γ, IL-6, tumor necrosis factor (TNF)-α, granulocyte macrophage colony-stimulating factor (GM-CSF) and IL-8 were observed in the postmortem brain tissues of ASD patients 22 , and another study found elevated levels of IFN-γ, monocyte chemoattractant protein (MCP)-1, IL-8, leukemia inhibitory factor (LIF) and interferon-γ-induced protein (IP)-10 23 . These widespread changes suggest that the cytokine signaling in ASD can be better characterized by a multivariate pattern of cytokines. There are reports in the literature that there are many associations between the levels of cytokines (such as MCP-1, IL-1β, IL-4, IL-6, etc.) and the main symptoms and adaptive functions of ASD children 24-26 . Therefore, there is a view that cytokines can be used as biomarkers to identify different subgroups of ASD. In each subgroup, ASD patients may have common immune-related pathologies and thus may have similar characteristics of treatment response 27 . Based on the above research findings, we analyzed the data obtained from the Shanghai Xinhua ASD Registry in China since 2016 to verify the following hypothesis: the immune activity of patients may help to identify ASD patients who respond best to bumetanide. Summary of the Invention

[0007] One aspect of the present invention provides a method for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, the method comprising:

[0008] obtaining characteristic information of the subject, wherein the characteristic information includes: (i) the baseline expression levels of a group of cytokines in the subject; (ii) the baseline behavioral performance of the subject; and (iii) the clinical information of the subject, including gender and age;

[0009] predicting the response of the subject to bumetanide based on the characteristic information.

[0010] In some embodiments, the cytokine group includes three or more cytokines selected from the group consisting of IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ.

[0011] In some embodiments, the cytokine group includes a group selected from:

[0012] (i) IL16, GROα, and IL7;

[0013] (ii) IL16, GROα, and TNFβ;

[0014] (iii) IL16, GROα, IL7, TNFβ, and CTACK; and

[0015] (iv) IL16, GROα, IL7, TNFβ, LIF, and MIF.

[0016] In some embodiments, the cytokine group includes IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ.

[0017] In some embodiments, the said behavioral performance includes the scores of one or more ASD screening tools or diagnostic tools.

[0018] In some embodiments, the behavioral performance includes 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, or 13 scores selected from the group consisting of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN.

[0019] In some embodiments, the behavioral performance includes a set of scores selected from:

[0020] (i) ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D;

[0021] (ii) ADOS_S, ADOS_C, ADOS_P, and CARS_total;

[0022] (iii) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT;

[0023] (iv) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; and

[0024] (v) ADOS_S, ADOS_C, ADOS_P, and CARS_total.

[0025] In some embodiments, the method includes using a classifier to predict a subject's response to bumetanide based on feature information.

[0026] In some embodiments, the classifier is selected from the group consisting of oblique random forest (ORF), partial least squares (PLS), sparse linear discriminant analysis (sLDA), neural network (NN), and support vector machine (SVM).

[0027] In some embodiments, the classifier has been trained.

[0028] In some embodiments, the feature information includes: (i) the baseline expression levels of IL16, GROα, and IL7; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; (iii) gender and age; and the classifier is a support vector machine;

[0029] The characteristic information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age; the classifier is partial least squares method;

[0030] The characteristic information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, and CTACK; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; (iii) gender and age; the classifier is a neural network;

[0031] The characteristic information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, LIF, and MIF; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; (iii) gender and age; the classifier is sparse linear discriminant analysis; or

[0032] The characteristic information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age; the classifier is oblique random forest.

[0033] In some embodiments, the method includes obtaining characteristic information of a subject by measuring the expression levels of cytokines in a sample from the subject.

[0034] In some embodiments, the sample is plasma.

[0035] Another aspect of the present invention provides a method for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, the method comprising:

[0036] A classifier is trained using a training data set that includes the characteristic information of multiple ASD individuals and the determined responses of the individuals to bumetanide, and a subset of relevant characteristic information is selected, wherein the characteristic information includes: (i) the baseline expression levels of cytokines such as IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ; (ii) the baseline scores of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN; and (iii) gender and age.

[0037] Predict the response of a subject to bumetanide based on the subset of relevant characteristic information of the subject.

[0038] In some embodiments, the classifier is selected from the group consisting of oblique random forest (ORF), partial least squares (PLS), sparse linear discriminant analysis (sLDA), neural network (NN), and support vector machine (SVM).

[0039] In some embodiments, the expression level of the cytokine is obtained by measuring the expression level of the cytokine in a sample from the subject.

[0040] In some embodiments, the sample is plasma.

[0041] Another aspect of the present invention provides a method for treating autism spectrum disorder (ASD) in a subject, the method comprising:

[0042] Predict the response of a subject with ASD to bumetanide using any of the above methods;

[0043] Administer an effective amount of bumetanide to the subject identified as responsive to bumetanide.

[0044] Another aspect of the present invention provides a prediction device, comprising:

[0045] An input module for receiving characteristic information of a patient with autism spectrum disorder (ASD), wherein the characteristic information includes: (i) the baseline expression levels of a set of cytokines in the patient; (ii) the baseline behavioral manifestations of the patient; and (iii) the clinical information of the patient, including gender and age;

[0046] A classification module including a classifier, wherein the classifier is capable of predicting the response of a subject to bumetanide based on the characteristic information using the classifier.

[0047] In some embodiments, the prediction device further includes a training module configured to train the classifier using a training data set.

[0048] Another aspect of the present invention provides a computer-readable medium including computer-executable instructions recorded thereon for performing operations including:

[0049] Receiving characteristic information of a subject with autism spectrum disorder (ASD), wherein the characteristic information includes: (i) the baseline expression levels of a set of cytokines of the subject; (ii) the baseline behavioral manifestations of the subject; and (iii) the clinical information of the subject, including gender and age;

[0050] Predicting the response of the subject to bumetanide using a classifier algorithm based on the characteristic information.

[0051] In some embodiments, the operation further includes training the classifier using a training data set.

[0052] In some embodiments, the cytokine group includes three or more cytokines selected from the group consisting of IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ.

[0053] In some embodiments, the cytokine group includes a group selected from:

[0054] (i) IL16, GROα, and IL7;

[0055] (ii) IL16, GROα, and TNFβ;

[0056] (iii) IL16, GROα, IL7, TNFβ, and CTACK; and

[0057] (iv) IL16, GROα, IL7, TNFβ, LIF, and MIF.

[0058] In some embodiments, the cytokine panel includes IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ.

[0059] In some embodiments, the behavioral manifestations include 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, or 13 scores selected from the group consisting of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN.

[0060] In some embodiments, the behavioral manifestations include a set of scores selected from the following:

[0061] (i) ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D;

[0062] (ii) ADOS_S, ADOS_C, ADOS_P, and CARS_total;

[0063] (iii) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT;

[0064] (iv) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; and

[0065] (v) ADOS_S, ADOS_C, ADOS_P, and CARS_total.

[0066] In some embodiments, the classifier is selected from the group consisting of oblique random forest (ORF), partial least squares (PLS), sparse linear discriminant analysis (sLDA), neural network (NN), and support vector machine (SVM).

[0067] In some embodiments, the classifier has been trained.

[0068] In some embodiments, the feature information includes: (i) the baseline expression levels of IL16, GROα, and IL7; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; (iii) gender and age; and the classifier is a support vector machine.

[0069] The feature information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age; and the classifier is partial least squares.

[0070] The feature information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, and CTACK; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; (iii) gender and age; and the classifier is a neural network.

[0071] The feature information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, LIF, and MIF; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; (iii) gender and age; and the classifier is sparse linear discriminant analysis; or

[0072] The feature information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age; and the classifier is oblique random forest.

[0073] Another aspect of the present invention provides a kit for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, the kit comprising reagents for measuring the expression levels of any one of the above-described sets of cytokines and instructions describing any one of the above-described prediction methods.

[0074] Another aspect of the present invention provides a kit for treating autism spectrum disorder (ASD) in a subject, the kit comprising:

[0075] An agent for measuring the expression level of any one of the above groups of cytokines;

[0076] Bumetanide; and

[0077] A specification describing any one of the above prediction methods, and administering an effective amount of bumetanide to a subject identified as responsive to bumetanide.

[0078] Another aspect of the present invention provides the above-described characteristic information for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide.

[0079] Another aspect of the present invention provides using the above-described characteristic information to prepare a set of characteristics for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide.

[0080] Another aspect of the present invention provides the use of an agent for measuring the expression level of any one of the above groups of cytokines in the preparation of an agent or a kit for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, wherein the prediction is performed by any one of the above prediction methods.

[0081] Another aspect of the present invention provides establishing a prediction model using the above-described characteristic information to predict the response of a patient with autism spectrum disorder (ASD) to bumetanide. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 . Sparse canonical correlation analysis. We performed sparse canonical correlation analysis on the [A] discovery set and the [B] validation set. The canonical scores between CARS and cytokines were min-max standardized and log-transformed, and were significantly correlated in both datasets.

[0083] Figure 2 . Differences among three immune behavior groups. K-means clustering map on the immune behavior plane. K-means clustering analysis was performed in the [A] discovery set, and the patients in the validation set were mapped to this immune behavior plane [B]. [C] Radar chart of the change ratios of CARS and cytokine levels relative to the baseline in three immune behavior groups. [D] Box plot of the significant changes in CARS and cytokine levels in three immune behavior groups; the box plots from left to right are the best response group, the worst response group, and the medium response group, respectively.

[0084] Figure 3.ROC curves for predicting responder groups defined by immune behavior. The classifiers include an oblique random forest (ORF) model, a partial least squares (PLS) model, a support vector machine (SVM) model, a sparse linear discriminant analysis (sLDA) model, and a neural network (NN) model. The models were trained based on the immune-behavior covariation plane to predict the response of autistic children to bumetanide. As described in the main text, the models were trained using the discovery set and tested using the validation set. The classification accuracy of the test dataset is shown in the figure below. [A] Model for predicting the baseline cytokine levels of ASD patients in the best responder group. [B] Model for predicting the ASD patients in the best responder group without baseline cytokine levels. [C] Model for predicting the baseline cytokine levels of ASD patients in the worst responder group. [D] Model for predicting the baseline cytokine levels of ASD patients in the worst responder group.

[0085] Figure 4 .Adjusting batch effects. The PCA plots show the effect of the "ComBat" algorithm adjustment on the batch effects of two batches. Scatter plots of the first two principal components (PC1 vs PC2) are shown before and after ComBat adjustment. In these plots, individual patient samples are represented as points and color-coded according to their source batch. After ComBat adjustment, the batches showed higher homogeneity, as demonstrated by the increased overlap of the PCA scatter plots.

[0086] Figure 5 .Pairwise associations between the total CARS score and cytokines. The partial correlation heatmap shows the pairwise associations between [A] the baseline total CARS score and baseline cytokine levels, [B] the baseline total CARS score and changes in cytokine levels, [C] changes in the CARS total score and baseline cytokine levels, and [D] changes in the CARS total score and changes in cytokine levels. The associations were adjusted for FDR.

[0087] Figure 6 .Scree plot for selecting the optimal number of clusters. We generated different clusters with the number of clusters (k) ranging from 2 to 14. We employed internal cluster quality measurement methods to validate the clustering results. For each possible number of clusters, an index was calculated to reflect the similarity between topics within the cluster and the dissimilarity between clusters. This index generally increases monotonically with the increase in the number of clusters, and the optimal value is determined as the elbow of its plot, i.e., the place where the change in the index (the difference from k - 1 and k + 1) is the largest. There are several possible indices available. We used the number of clusters selected by most people. The figure below shows the Hilbert statistic for k possible clusters (from 2 to 14 in total) and the change in the index (increment) compared to k - 1. The results of the internal cluster quality index produced a good signal in 3 clusters.

[0088] Figure 7. Box plots of the changes in CARS and cytokine levels in the three immune behavior groups. The boxes from left to right represent the best response group, the worst response group, and the moderate response group

[0089] Figure 8 . ROC curves defined by CARS for predicting treatment response. The classifiers include the oblique random forest (ORF) model, partial least squares (PLS) model, support vector machine (SVM) model, sparse linear discriminant analysis (sLDA) model, and neural network (NN) model. The models were trained based on the behavioral assessment at the pre-treatment baseline to predict the response of ASD children to bumetanide. After 3 months of bumetanide treatment, a responder was defined as a child whose total CARS score decreased by more than 2.5 points (left) or 2 points (right). As described in the example, the models were trained using the discovery set and tested using the validation set. The classification accuracy of the test dataset is shown in the figure.

[0090] Figure 9 . ROC curves for predicting the best treatment response defined by CARS. The five classifiers (from left to right) include the support vector machine (SVM) model, partial least squares (PLS) model, neural network (NN) model, sparse linear discriminant analysis (sLDA) model, and oblique random forest (ORF) model.

[0091] Detailed Description

[0092] Bumetanide is a drug being studied for the treatment of autism spectrum disorder (ASD). It may act to restore the function of gamma-aminobutyric acid (GABA), which may be regulated by the immune system. However, the interaction between bumetanide and the immune system remains unclear. The treatment of 79 children with ASD with bumetanide for 3 months was analyzed from a longitudinal sample. The covariation between symptom improvement and cytokine changes was calculated and validated by sparse canonical correlation analysis. The response patterns to bumetanide were revealed by cluster analysis. Five classifiers were used to test whether including baseline information on cytokines could improve the prediction of response patterns using an independent test sample. There was an immunological-behavioral covariation between symptom improvement in the Childhood Autism Rating Scale (CARS) and cytokine changes between interferon (IFN)-γ, monokine induced by gamma interferon, and IFN-α2. Three groups with different response patterns to bumetanide were detected using this covariation, including the best-response group (21.5%, n = 17; Hedge improvement g = 2.16 in the CARS), the worst-response group (22.8%, n = 18; g = 1.02), and the moderate-response group (55.7%, n = 44; g = 1.42). The model including cytokine levels significantly improved the prediction of the best-response group before treatment (optimal area under the curve, AUC = 0.832) compared with the model not including cytokine levels (95% confidence interval for the improvement in AUC was [0.287, 0.319]). Cytokine measurement helps identify potential responders to bumetanide in children with ASD, suggesting that the immune response may interact with the mechanism of action of bumetanide to enhance GABA function in ASD.

[0093] It should be understood that the specific methods and conditions described in the embodiments of the present invention are only for describing the specific embodiments and are not intended to limit the present invention, and any methods and conditions similar or equivalent to those described herein can be adopted in the practice or testing of the present invention. The explanations of the theories or mechanisms in the present invention are only used to help understand the present invention and should not be regarded as a limitation to the embodiments protected by the present invention.

[0094] Unless otherwise specified, the terms used in the present invention have the meanings commonly understood in the art and can be understood by referring to standard textbooks, references, and literature known to those skilled in the art. All publications mentioned herein are incorporated herein by reference in their entirety.

[0095] The term

[0096] It must be noted that, as used in this specification and the appended claims, the singular forms "a", "an", and "the" include plural referents unless the context clearly dictates otherwise. It should also be noted that the claims may be drafted to exclude any optional element. Thus, this statement is intended to serve as antecedent basis for using exclusive terms, such as "only", "solely", etc. or using "negative" limitations in reciting claim elements.

[0097] Unless otherwise indicated, the terms "comprise", "comprising", "include", and variations thereof, such as "includes" and "including", are not intended to exclude other components, elements, wholes, or steps. These terms also encompass the meaning of "consisting of" or "consisting essentially of". The term "consisting of" or "consisting essentially of" is a specific embodiment of the term "comprise", wherein any other unrecited component, element, whole, or step is excluded.

[0098] The term "about" refers to a range that is equal to a specific value plus or minus ten percent (+ / - 10%).

[0099] The term "and / or" refers to any one, several, or all of the elements connected by that term.

[0100] As used herein, the term "expression" or "expressed" refers to the translation of an RNA molecule into a protein, polypeptide, or a portion thereof.

[0101] As used herein, the term "biomarker" refers to a gene or protein or a combination of multiple genes or multiple proteins, the expression or concentration level of which in a sample is altered after treatment with a therapeutic agent (e.g., bumetanide) or indicates that a disease (e.g., ASD) responds to the therapeutic agent (e.g., bumetanide). The biomarkers disclosed herein are genes and / or proteins whose expression level or concentration, or change in expression or concentration, is related to the responsiveness of ASD to bumetanide.

[0102] The terms "subject" and "patient" may be used interchangeably herein. As used herein, the term "subject" refers to any organism to which the methods of the present invention can be applied, e.g., for experimental, diagnostic, prophylactic, and / or therapeutic purposes. Typical subjects include animals (e.g., mammals such as mice, rats, rabbits, non-human primates such as chimpanzees and other apes and monkeys, and humans). The subject may be a mammal, particularly a human, including male or female, and including neonates, infants, juveniles, adolescents, adults, or the elderly, and also including various races and ethnicities. In some instances, a subject refers to an individual in need of diagnosis, treatment, or prophylaxis of a disease or disorder, who may have the disease or disorder or be at risk of developing the disease or disorder.

[0103] As used herein, the term "sample" refers to any sample of cells, tissues, or body fluids from which biomarker expression can be detected. Examples of such samples include, but are not limited to, biopsy samples, smears, blood, lymph, urine, saliva, or any other bodily secretion or derivative thereof. For example, blood can include whole blood, plasma, serum, or any blood derivative. Samples can be obtained from a subject by a variety of techniques known to those of skill in the art.

[0104] As used herein, the term "treatment" refers to both therapeutic treatment and prophylactic or preventive measures, the purpose of which is to prevent or slow down (mitigate) a target pathological condition or disorder. Persons in need of treatment include those diagnosed with the disease, those predisposed to the disease (e.g., those with a genetic predisposition), or those in need of preventing the disease. The term "prevention" refers to reducing the likelihood of the occurrence (or recurrence) of a disease, disorder, condition, or related symptoms.

[0105] As used herein, the term "response" or "reactivity" refers to the effectiveness of a treatment or therapy in alleviating a disease or reducing symptoms. Any endpoint indicating a benefit to the subject can be used to evaluate a beneficial response, including but not limited to: (1) inhibiting disease progression to some extent, including slowing and completely stopping; (2) improving (e.g., reducing the number, frequency, and / or intensity) of one or more disease symptoms; (3) stabilizing the condition, e.g., preventing or delaying deterioration that would be expected or normally observed in the absence of treatment; etc. A beneficial response to treatment in an ASD patient may manifest as an improvement in one or more behaviors associated with ASD, such as those listed in common screening tools or diagnostic tools.

[0106] The term "classification" refers to a procedure and / or algorithm in which individual items are placed into groups or classes based on quantitative information about one or more inherent characteristics (referred to as features, variables, characteristics, traits, etc.) of the items and a training set of previously labeled items based on a statistical model and / or.

[0107] For the methods of the present invention, the term "administering" refers to methods for prophylactically or therapeutically preventing, treating, or improving a syndrome, disorder, or disease (e.g., ASD) as described herein. The methods include administering an effective amount of a therapeutic agent (e.g., bumetanide) during a treatment course. The mode of administration should be understood to include all known suitable treatment regimens.

[0108] The term "pharmaceutically acceptable salt" as used herein refers to relatively non-toxic inorganic or organic acid salts of the compounds of the present invention. These salts can be prepared in situ during the final isolation and purification of the compound, or by reacting the purified compound in its free form with a suitable organic or inorganic acid separately and isolating the salt thus formed. Representative acid salts include, but are not limited to, acetate, adipate, aspartate, benzoate, benzenesulfonate, bicarbonate / carbonate, bisulfate / sulfate, borate, camphorsulfonate, citrate, cyclohexylsulfamate, ethanedisulfonate, esylate, formate, fumarate, glucoheptonate, gluconate, glucuronate, hexafluorophosphate, hippurate, hydrochloride / chloride, hydrobromide / bromide, hydroiodide / iodide, hydroxyethylsulfonate, lactate, malate, maleate, malonate, mesylate, methylsulfate, naphthoate, 2-naphthalenesulfonate, nicotinate, nitrate, orotate, oxalate, palmitate, pamoate, phosphate / hydrogenphosphate / dihydrogenphosphate, pyroglutamate, saccharate, stearate, succinate, tannate, tartrate, mesylate, trifluoroacetate, and cinchomeronate. In one embodiment, the pharmaceutically acceptable salt is a hydrochloride / chloride salt.

[0109] The term "solvate" as used herein refers to a variable stoichiometric complex formed by a solute (e.g., the active agent of the present invention) and a solvent. For the purposes of the present invention, such a solvent shall not interfere with the biological activity of the solute. Examples of suitable solvents include, but are not limited to, water, methanol, ethanol, and acetic acid.

[0110] The term "tautomer" as used herein refers to two (or more) compounds that differ only in the position and electronic distribution of one (or more) mobile atoms, such as keto-enol tautomers and imine-enamine tautomers.

[0111] The term "effective amount" or "therapeutically effective amount" refers to the amount of the active compound or agent provided herein, a combination of therapeutic compounds, or a pharmaceutical composition thereof that elicits a biological or medical response in an animal or human tissue system sought by a researcher, veterinarian, physician, or other clinician, including preventing, treating, or ameliorating the syndrome, disorder, or disease being treated, or the symptoms of the syndrome, disorder, or disease being treated (e.g., ASD).

[0112] The term "computer" as used herein includes at least one hardware processor using at least one memory. The at least one memory can store a set of instructions. The instructions can be stored permanently or temporarily in the memory of the computer. The processor executes the instructions stored in the memory to process data. The set of instructions can include various instructions for performing a particular task or multiple tasks, such as the tasks described herein.

[0113] Prediction of Response to Bumetanide in Autism Spectrum Disorder (ASD)

[0114] Autism spectrum disorder (ASD) is a clinically diagnosed disorder with single and complex polygenic etiologies and can be diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5). Autism spectrum disorder is characterized by varying degrees of impairment in communication skills and social interactions, as well as restrictive, repetitive, and stereotyped patterns of behavior, including autism, pervasive developmental disorder not otherwise specified (PDD-NOS), Asperger syndrome, Rett syndrome, and childhood disintegrative disorder (CDD). The ASD described in the present invention may also refer to any of the above specific disorders or syndromes. In some embodiments, ASD may be a childhood autism spectrum disorder.

[0115] A variety of tools can be used to screen for and diagnose ASD, such as various scales or subscales designed for screening or diagnosing ASD. Some common screening tools can assist in screening for and diagnosing ASD, including the Checklist for Autism in Toddlers (CHAT), Modified Checklist for Autism in Toddlers (M-CHAT), Screening Tool for Autism in Two-Year-Olds (STAT), Social Communication Questionnaire (SCQ) (for children 4 years of age and older), Autism Spectrum Screening Questionnaire (ASSQ), Australian Asperger's Syndrome Scale, and Children's Asperger's Syndrome Test (CAST).

[0116] Typical diagnostic tools include, but are not limited to, the Autism Diagnostic Interview-Revised (ADI-R), Autism Diagnostic Observation Schedule (ADOS), Childhood Autism Rating Scale (CARS), Autism Behavior Checklist (ABC), and Social Responsiveness Scale (SRS). These tools can be used to test behavioral manifestations, and those skilled in the art know how to conduct behavioral tests and obtain scores for these tools. For example, the scores for each item in a scale or subscale can be added together to obtain the score for that scale or subscale.

[0117] The CARS consists of 15 items and is scored on a 7-point scale (ranging from 1 to 4); the higher the score, the higher the degree of impairment (Schopler E, Reichler R, DeVellis R. Toward objective classification of childhood autism: Childhood Autism Rating Scale (CARS). Journal of Autism and Developmental Disorders. 1980;10:91-103; Schopler E, Reichler R, RochenRenner B. Childhood Autism Rating Scale. Western Psychological Services; 1988). The total score ranges from a minimum of 15 to a maximum of 60; a score below 30 indicates that the individual is in the non-autistic range, a score between 30 and 36.5 indicates mild to moderate autism, and a score between 37 and 60 indicates severe autism. The items in the CARS can be further divided into three subscales: Social Impairment (SI), Negative Emotions (NE), and Distorted Sensory Responses (DSR) (DiLalla DL, Rogers SJ. Domains of the Childhood Autism Rating Scale: Relevance to diagnosis and treatment. Journal of Autism and Developmental Disorders 1994;24(2):115-128). In the present invention, the total number of items and the CARS scores of the three subscales SI, NE, and DSR are respectively designated as CAR_total, CAR_S, CAR_N, and CAR_D, as shown in Table 1.

[0118] The ADOS (also known as the Autism Diagnostic Observation Schedule - Generic, ADOS - G) is a semi-structured assessment of social interaction, communication, play, and creative use of materials for individuals who may have autism or an autism spectrum disorder (ASD). The ADOS consists of four "modules". Module 1 is for children who have not yet learned to speak or who can only use single words. Module 2 is for individuals with phrase-expressive abilities. Module 3 is used for children and adolescents with fluent verbal expression. Module 4 is used to evaluate adolescents and adults with fluent speech (Lord et al., Autism Diagnostic Observation Schedule - Generic: A standard measure of social and communication deficits associated with the autism spectrum, Journal of Autism and Developmental Disorders, 2000, 30(3):205-223; Lord, C., Rutter, M., DiLavore, P.C., & Risi, S. (2008). Manual for the Autism Diagnostic Observation Schedule. Los Angeles: Western Psychological Services). In the present invention, the ADOS scores for the subscales of social, communication, play, and creative use of materials are respectively designated as ADOS_S, ADOS_C, ADOS_P, and ADOS_I, as shown in Table 1. The higher the score, the more severe the autism symptoms.

[0119] The Social Responsiveness Scale (SRS) is a 65-item questionnaire that is a standardized measure of the core symptoms of autism. Each item is rated on a 4-point Likert scale. The scores of each individual item are summed to obtain a total raw score. A total score of 0 - 62 is in the normal range, a total score of 63 - 79 indicates mild symptoms, a total score of 80 - 108 indicates moderate symptoms, and a total score of 109 - 149 indicates severe symptoms. Five subscales are also provided: Awareness of Social Awareness (AWA), Social Cognition (COG), Social Communication (COM), Social Motivation (MOT), and Mannerisms of Autism (MANN). (Constantino JN, Gruber CP. Social Responsiveness Scale: SRS-2 Western Psychological Services, Torrance, CA, 2012.) In the present invention, the SRS scores of the total items and the AWA, COG, COM, MOT, and MANN subscales are respectively designated as SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN, as shown in Table 1.

[0120] The Autism Diagnostic Interview - Revised (ADI-R) is a standardized, semi-structured, clinician-administered parent interview (Lord et al., Autism Diagnostic Interview - Revised: A revised diagnostic interview for caregivers of individuals with possible pervasive developmental disorders, Journal of Autism and Developmental Disorders, 1994, 24(5):659 - 685). The ADI-R includes 93 items that focus on early development, language / communication, reciprocal social interaction, and restricted, repetitive behaviors and interests. The higher the score, the more severe the autism symptoms.

[0121] The Aberrant Behavior Checklist (ABC) is a symptom rating inventory used to assess and classify problem behaviors in children and adults in a variety of settings (Aman et al., Psychometric characteristics of the Aberrant Behavior Checklist. American Journal of Mental Deficiency, March 1985; 89(5):492 - 502). The ABC includes 58 items that are divided into five subscales: (1) Irritability, (2) Lethargy / Social Withdrawal, (3) Stereotypy, (4) Hyperactivity / Noncompliance, and (5) Inappropriate Speech. The higher the score, the more severe the autism symptoms.

[0122] It has been reported that bumetanide can improve the main symptoms of ASD, but only a part of ASD patients can benefit from bumetanide treatment. The present inventors have found that cytokines can be used to evaluate and predict the response of a subject to bumetanide, that is, to evaluate and predict the therapeutic effect of bumetanide on ASD patients. The age of the subject can be in the range of about 1 to about 45 years old, about 2 to about 40 years old, about 3 to about 30 years old, about 3 to about 20 years old, about 3 to about 12 years old, about 3 to about 10 years old. The subject can be male or female. In some embodiments, the subject may be a child with ASD. In some embodiments, the child is 3 to 12 years old. In some embodiments, the child is 3 to 10 years old.

[0123] As used herein, the term "bumetanide" refers to 3-butylamino-4-phenoxy-5-sulfamoylbenzoic acid, or a pharmaceutically acceptable salt, solvate or tautomer thereof.

[0124] According to the present invention, the characteristic information of a subject with ASD can be used to predict the response of the subject to bumetanide treatment. The characteristic information can include various characteristics of the subject, and can also be interpreted as a set of multiple characteristics of the subject. The characteristic information can include: (i) the baseline expression level of a group of cytokines in the subject; (ii) the baseline behavioral performance of the subject; and (iii) the clinical information of the subject, including gender, age, and BMI, etc.

[0125] The term "baseline", when used in combination with characteristic information, refers to the characteristic information of the subject before receiving bumetanide treatment. The baseline characteristic information can be measured or evaluated at any appropriate time before the subject is treated with bumetanide, such as half a day to one week before the subject is treated with bumetanide. It should be understood that different characteristic information can be obtained at different time points within an appropriate time period. The baseline expression level refers to the cytokine expression level of the subject before the subject receives bumetanide treatment. The baseline behavioral performance refers to the behavioral performance of the subject before the subject receives bumetanide treatment.

[0126] The set of cytokines can include 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48 or more cytokines. In some embodiments, the set of cytokines can be selected from the cytokines listed in Table 1, namely IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGF-BB, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9 and SCGFβ. In some embodiments, the set of cytokines can include any 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34 or all 35 cytokines listed in Table 1.

[0127] It should be understood that the cytokines listed in Table 1 are merely illustrative and are not intended to limit the scope of the present invention. The inventors have found that the baseline expression levels of cytokines are correlated with the response of ASD subjects to bumetanide, and the cytokines described herein can include any type and any number of cytokines and are not limited to the cytokines listed in Table 1. Those skilled in the art can identify specific cytokines for predicting the response of ASD subjects to bumetanide through different algorithms or using different prediction models (as described below). This group of cytokines can include other cytokines not listed in Table 1, that is, this group of cytokines can include 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48 or more cytokines, some of which are cytokines selected from the cytokines listed in Table 1 and others are cytokines not listed in Table 1. In some embodiments, this group of cytokines can include 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47 or 48 cytokines selected from basic FGF (bFGF), β-NGF CTACK, eosinophil chemotactic factor, G-CSF, GM-CSF, GRO-α, HGF, IFN-α2, IFN-γ, IL-10, IL-12p40, IL-12p70, IL-13, IL-15, IL-16, IL-17, IL-18, IL-1α, IL-1β, IL-1Ra, IL-2, IL-2Rα, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IP-10, LIF, M-CSF, MCP-1, MCP-3, MIF, MIG, MIP-1α, MIP-1β, PDGF-BB, RANTES, SCF, SCGF-β, SDF-1α, TNF-α, TNF-β, TRAIL and VEGF.

[0128] In some embodiments, this group of cytokines includes groups selected from:

[0129] (i) IL16, GROα;

[0130] (ii) IL16, GROα and IL7;

[0131] (iii) IL16, GROα, and TNFβ;

[0132] (iv) IL16, GROα, IL7, TNFβ, and CTACK;

[0133] (v) IL16, GROα, IL7, TNFβ, LIF, and MIF.

[0134] The expression level of cytokines can be obtained by measuring the cytokine concentration in a sample from a subject. In some embodiments, the sample can be blood (e.g., peripheral blood), plasma, or serum.

[0135] The expression level of the cytokines can be measured by a variety of methods known in the art. The methods include, for example, immunoassay, radioimmunoassay (RIA), immunoradiometric assay (IRMA), enzyme-linked immunosorbent assay (ELISA), Western blot analysis, ELISpot, CELISA (cellular enzyme-linked immunosorbent assay), RHPA (reverse hemolytic plaque assay), kinase receptor activation assay (KIRA), cytokine immunocapture assay (CITA), or radioligand receptor assay (RRA). The assays can be performed in a multiplex or matrix-based format (e.g., multiplex immunoassay). In some embodiments, the level of the cytokines can be determined by bead-based multiplex assays, such as the Bio-Plex multiplex immunoassay system.

[0136] Behavioral performance can be evaluated or further quantified by any method known in the art. Behavioral performance can be measured by the scores of any of the above screening tools or diagnostic tools, such as CARS, ADOS, SRS, ADI-R, or ABC, or any of their subscales, or any combination of them and their subscales, as shown in Table 1.

[0137] In some embodiments, the behavioral performance includes the CARS scores, namely CAR_total, CAR_S, CAR_N, CAR_D, or any combination thereof. In some embodiments, the behavioral performance includes CAR_total, CAR_S, CAR_N, and CAR_D. In some embodiments, the behavioral performance includes any 1, 2, 3, or 4 selected from the group consisting of CAR_total, CAR_S, CAR_N, and CAR_D.

[0138] In some embodiments, the behavioral manifestations include ADOS scores, namely ADOS_S, ADOS_C, ADOS_P, ADOS_I, or any combination thereof. In some embodiments, the behavioral manifestations include ADOS_S, ADOS_C, ADOS_P, and ADOS_I. In some embodiments, the behavioral manifestations include any 1, 2, 3, or 4 selected from the group consisting of ADOS_S, ADOS_C, ADOS_P, and ADOS_I.

[0139] In some embodiments, the behavioral manifestations include SRS scores, namely SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, SRS_MANN, or any combination thereof. In some embodiments, the behavioral manifestations include SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN. In some embodiments, the behavioral manifestations include any 1, 2, 3, 4, or 5 selected from the group consisting of SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN.

[0140] In some embodiments, the behavioral manifestations include CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN. In some embodiments, the behavioral manifestations include any 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, or 13 selected from the group consisting of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN.

[0141] In some embodiments, the behavioral manifestations include groups selected from the following:

[0142] (i) ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D;

[0143] (ii) ADOS_S, ADOS_C, ADOS_P, and CARS_total;

[0144] (iii) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT;

[0145] (iv) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; and

[0146] (v) ADOS_S, ADOS_C, ADOS_P, and CARS_total.

[0147] The clinical information of the subject includes gender, age, BMI, etc. In some embodiments, the clinical information includes gender and age.

[0148] In some embodiments, the characteristic information for predicting the response of a subject to bumetanide treatment may include: (i) the baseline expression levels of cytokines such as IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ; (ii) the baseline scores of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN; and (iii) gender and age.

[0149] In some embodiments, the characteristic information for predicting the response of a subject to bumetanide treatment may include: (i) the baseline expression levels of IL16, GROα, and IL7; (ii) the baseline scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; and (iii) gender and age.

[0150] In some embodiments, the characteristic information for predicting the response of a subject to bumetanide treatment may include: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the baseline scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; and (iii) gender and age.

[0151] In some embodiments, the characteristic information for predicting a subject's response to bumetanide treatment may include: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, and CTACK; (ii) the baseline scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; and (iii) gender and age.

[0152] In some embodiments, the characteristic information for predicting a subject's response to bumetanide treatment may include: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, LIF, and MIF; (ii) the baseline scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; and (iii) gender and age.

[0153] In some embodiments, the characteristic information for predicting a subject's response to bumetanide treatment may include: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the baseline scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; and (iii) gender and age.

[0154] The response of a subject to bumetanide can be interpreted as the therapeutic effect of bumetanide on the subject. In the present invention, the response or therapeutic effect may refer to the response or therapeutic effect exhibited by the subject after receiving bumetanide treatment for a treatment period. The treatment period is sufficient for bumetanide to show its therapeutic effect in the subject. The treatment duration may be about 1 month to 6 months, such as 2 to 3 months, 2 to 4 months.

[0155] The response of a subject to bumetanide can be classified as responsive or non-responsive, highly responsive or lowly responsive, better responsive or worst responsive, positive responsive or negative responsive, or responsive or non-responsive, etc., according to known criteria or criteria determined by statistical analysis of a group of ASD patients receiving bumetanide treatment. Generally, a good therapeutic effect is indicated by being responsive, highly responsive, better responsive, positive responsive, or a responder.

[0156] In some embodiments, response can be determined by the change in the score of a screening or diagnostic tool (such as the aforementioned tool) after treatment with bumetanide relative to the score before treatment. Typically, a decrease in the score (such as the CARS score, ADOS score, or SRS score) above a particular threshold indicates improvement in ASD, while a decrease in the score (such as the CARS score, ADOS score, or SRS score) below a particular threshold, or an increase or no change in the score, indicates no improvement in ASD. For common screening or diagnostic tools (such as the aforementioned tool), the threshold is known to those skilled in the art. "Responsive", "highly responsive", "better responsive", "positively responsive", or "responder" to bumetanide can refer to a decrease in the score of a screening or diagnostic tool (such as the CARS score, e.g., CARS_total) after treatment with bumetanide relative to before treatment that is not less than (e.g., higher than or greater than) a particular threshold. The specific threshold for "responsive", "highly responsive", "better responsive", "positively responsive", or "responder" may be about 2, 2.5, or 3. "Non-responsive", "low responsive", "least responsive", "negatively responsive", or "non-responder" to bumetanide can refer to a decrease in the score of a screening or diagnostic tool (such as the CARS score, e.g., CARS_total) after treatment with bumetanide relative to before treatment that is less than (or smaller than) a particular threshold, or a score of a screening or diagnostic tool (such as the CARS score, e.g., CARS_total) after treatment with bumetanide that is higher than or substantially equal to the score before treatment. For example, CARS_total after treatment with bumetanide is higher than or substantially equal to that before treatment, or is lower than that before treatment by a value less than or equal to about 1 or 2. The specific threshold for "non-responsive", "low responsive", "least responsive", "negatively responsive", or "non-responder" may be 1 or 2.

[0157] In some embodiments, "responsive", "highly responsive", "better responsive", "positively responsive", or "responder" to bumetanide can be defined as determining better or best treatment efficacy in a population of ASD patients treated with bumetanide using statistical analysis. "Non-responsive", "low responsive", "least responsive", "negatively responsive", or "non-responder" to bumetanide can be defined as determining the worst treatment efficacy in a population of ASD patients treated with bumetanide using statistical analysis.

[0158] "Better", "best", or "worst" means that the treatment efficacy of one group of people is statistically significantly different from that of other groups of people. Treatment efficacy can be quantified by the change in the score of any screening tool or diagnostic tool, and generally, the greater the decrease in the score, the better the treatment efficacy. Those skilled in the art know how to judge treatment efficacy based on screening tools or diagnostic tools.

[0159] In some embodiments, the statistical analysis may be cluster analysis, including but not limited to centroid-based (e.g., k-means) clustering, hierarchical clustering (e.g., mean shift or agglomerative hierarchy), distribution-based clustering, density-based clustering (e.g., density-based spatial clustering of applications with noise, DBSCAN), grid-based clustering. "Responders", "high responders", "better responders", "positive responders" or "responders" to bumetanide may be defined as the responses of the clusters determined to have the best treatment effect using cluster analysis. "Non-responders", "low responders", "lowest responders", "negative responders" or "non-responders" to bumetanide may be defined as the responses of the clusters determined to have the least treatment effect using cluster analysis. In some embodiments, the cluster analysis is performed based on the changes in the screening or diagnostic tool scores (e.g., CARS score, such as, CARS_total) of individuals in the population after receiving bumetanide treatment compared to before treatment and the changes in the expression levels of cytokines (e.g., MIG, IFN-α2, IFN-γ).

[0160] In some embodiments, "responders", "high responders", "better responders", "positive responders" or "responders" to bumetanide may be defined as having a treatment effect superior to the third quartile in the population of ASD patients treated with bumetanide. "Non-responders", "low responders", "lowest responders", "negative responders" or "non-responders" to bumetanide may be defined as having a treatment effect worse than the first of the second quartile in the population of ASD patients treated with bumetanide. The quartiles may be determined based on the treatment effect of bumetanide on the population of ASD patients.

[0161] There is no mandatory requirement for the number of patients included in the population for determining the bumetanide response criteria, and statistical results can be obtained. For example, the number of patients included in the population may be 50 - 200, such as 50 - 100.

[0162] Predicting the response of autistic subjects to bumetanide is essentially a classification method, and the prediction result is the classification result, that is, classifying the subjects as having different responses to bumetanide. A prediction model can be used to predict the response of subjects with ASD to bumetanide. This prediction model can also be called a classifier and can be used to determine the response of a subject to bumetanide. The classifier may be a machine learning system and can characterize the response of ASD to bumetanide based on the characteristic information of the subject. In the present invention, using a classifier means that the characteristic information of the subject can be used as the input of the classifier, and the output is the response of the subject to bumetanide.

[0163] Typically, the model is constructed using the characteristic information of subjects for which the classification (i.e., the response of ASD patients to bumetanide) has already been determined. Once the model (classifier) is established, it can be applied to the characteristic information obtained from subjects with ASD to determine the response of the subjects to bumetanide.

[0164] A variety of prediction models known in the art can be used as the classifier of the present invention to determine the response of a subject to bumetanide. For example, the classifier may include an algorithm selected from a support vector machine (SVM), partial least squares (PLS), neural network (NN), sparse linear discriminant analysis (sLDA), oblique random forest (ORF), logistic regression, quadratic discriminant analysis (QDA), naive Bayes, C4.5 decision tree, or k-nearest neighbor (KNN).

[0165] The classifier may be a trained classifier. In some embodiments, the method of the present invention further includes the step of training the classifier before prediction. The classifier may have been trained using a training dataset to select the optimal classification algorithm and establish a prediction model. The training dataset may include the characteristic information of a plurality of individuals with ASD and the response of the individuals to bumetanide, which is determined, for example, after the individuals are treated with bumetanide for a period of time (e.g., one month to six months, such as two months to three months). The training dataset may also include control individuals who have been determined not to have ASD or who have been determined to have ASD but have not received bumetanide treatment (e.g., received placebo treatment).

[0166] The age range of the population in the training dataset may be from about 1 year to about 45 years, about 2 years to about 40 years, about 3 years to about 30 years, about 3 years to about 20 years, about 3 years to about 12 years, about 3 years to about 10 years. The average age of the population in the training dataset may be 3 years, 4 years, 5 years, 6 years, 7 years, 8 years, 9 years, 10 years or older. The population may consist entirely of males, or entirely of females, or may consist of both males and females.

[0167] Optionally, once the model is established, methods known in the art can be used to verify the effectiveness of the model. In some embodiments, the method of the present invention further includes the step of verifying the classifier after training and before prediction. One way to verify the effectiveness of the model is to apply the model to an independent dataset, such as a validation dataset. The validation dataset may include the characteristic information of multiple individuals with ASD and the determined response of the individuals to bumetanide, and the individuals in the validation dataset are different from those in the training dataset. Another way to verify the model is through cross-validation of the dataset. As described above, for cross-validation, one or a subset of individuals need to be excluded, and then the model is constructed. The individuals that do not need to be excluded form the "cross-validation model". As described herein, the individuals excluded according to the model prediction are determined. This process is completed for all individuals or subsets of the initial dataset, and the error rate is determined. Subsequently, the accuracy of the model is evaluated.

[0168] The age range of the population in the validation dataset can be from about 1 year to about 45 years old, from about 2 years to about 40 years old, from about 3 years to about 30 years old, from about 3 years to about 20 years old, from about 3 years to about 12 years old, from about 3 years to about 10 years old. The average age of the population in the validation dataset may be about 3 years old, 4 years old, 5 years old, 6 years old, 7 years old, 8 years old, 9 years old, 10 years old or older. The population can be all male, or all female, or can be composed of both male and female.

[0169] In some embodiments, the characteristic information of the individuals in the training dataset and the validation dataset may independently include any one of the following groups:

[0170] (i) The baseline expression levels of cytokines such as IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9 and SCGFβ; the baseline scores of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT and SRS_MANN; and gender and age;

[0171] (ii) Baseline expression levels of IL16, GROα, and IL7; baseline scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; and gender and age;

[0172] (iii) Baseline expression levels of IL16, GROα, and TNFβ; baseline scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; and gender and age;

[0173] (iv) Baseline expression levels of IL16, GROα, IL7, TNFβ, and CTACK; baseline scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; and gender and age;

[0174] (v) Baseline expression levels of IL16, GROα, IL7, TNFβ, LIF, and MIF; baseline scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; and gender and age; or

[0175] (vi) Baseline expression levels of IL16, GROα, and TNFβ; baseline scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; and gender and age. Those skilled in the art will understand that the strength of the model can be evaluated by various parameters, including but not limited to accuracy and AUC. The methods for calculating accuracy and AUC are known in the art and are described herein (see, for example, the Examples). The accuracy of the predictive classifier can be at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95% or higher. The accuracy of the predictive classifier can be approximately 60%-70%, 70%-80%, 80%-90%, or 90%-100%. The AUC of the predictive classifier can be at least 0.60, at least 0.65, at least 0.70, at least 0.75, at least 0.80, at least 0.85, at least 0.80, at least 0.95 or higher. The AUC range of the predictive classifier can be approximately 0.60-0.70, 0.70-0.80, 0.80-0.90, or 0.90-1.00.

[0176] In some embodiments, the feature information may contain a large number of features, some of which may have a high correlation with the prediction result or contribute more, while some features may have a low correlation with the prediction result or no correlation. Therefore, relevant subsets of feature information can be selected from the feature information through feature selection techniques to improve the computational efficiency and accuracy of the classifier, and then the features in the subset can be used for prediction.

[0177] Common feature selection techniques in the art mainly include filtering techniques that evaluate the correlation of features by observing the inherent attributes of the data, wrapper methods that embed model assumptions into the search for feature subsets, and embedded techniques that build the search for the optimal feature set into the classifier algorithm. These techniques are well-known to those skilled in the art.

[0178] In the present invention, a prediction model (classifier) can also be used to select a subset of relevant feature information.

[0179] In some embodiments, a training data set containing the feature information of multiple individuals is used to train a classifier to select a subset of the feature information features. The features in the subset are highly correlated with the prediction result or contribute more to the prediction result. Subsequently, the classifier can use the features in the subset as input for prediction. As is known to those skilled in the art, the selection of the subset can be automatically completed by the classifier.

[0180] Those skilled in the art can understand that the subsets of relevant feature information selected using different classifiers may be different, which depends on the algorithms used by the classifiers. For a specific classifier, higher accuracy can be obtained by using the most suitable subset.

[0181] In some embodiments, the training data set for selecting the subset may include the feature information of multiple individuals with ASD and the determined responses of the individuals to bumetanide, where the feature information includes: (i) the baseline expression levels of a group of cytokines in the individual; (ii) the baseline behavioral manifestations of the individual; and (iii) the clinical information of the individual;

[0182] Among them, the cytokine group includes the cytokines listed in Table 1, namely IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ;

[0183] The behavioral manifestations include the scores listed in Table 1, namely CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN; and

[0184] The clinical information includes gender and age.

[0185] In some embodiments, the subset of relevant feature information includes: (i) the baseline expression levels of IL16, GROα, and IL7; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; (iii) gender and age. The classifier is a support vector machine.

[0186] In some embodiments, the subset of relevant feature information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age. The classifier is partial least squares.

[0187] In some embodiments, the subset of relevant feature information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, and CTACK; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; (iii) gender and age. The classifier is a neural network.

[0188] In some embodiments, the subset of relevant feature information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, LIF, and MIF; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; (iii) gender and age. The classifier is sparse linear discriminant analysis.

[0189] In some embodiments, the subset of relevant feature information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age. The classifier is oblique random forest.

[0190] It should be understood that variations of the prediction method of the present invention are also included within the scope of protection of the present invention. For example, providing the characteristic information or a subset of relevant characteristic information described herein for predicting the response of ASD subjects to bumetanide. For example, providing a set of characteristics for predicting the response of subjects with ASD to bumetanide using the characteristic information or a subset of relevant characteristic information described herein. For another example, providing the use of an agent (e.g., an antibody capable of specifically binding to the cytokine) for measuring the expression level of a group of cytokines described in the present invention in the preparation of an agent or kit for predicting the response of ASD subjects to bumetanide, wherein the prediction can be carried out, for example, by the prediction method described in the present invention. For example, providing a prediction model for predicting the response of subjects with ASD to bumetanide using a subset of the characteristic information or relevant characteristic information described herein.

[0191] The prediction method of the present invention can be carried out by a computer.

[0192] The present invention also provides a prediction device, which can be a computer device, and the prediction device includes:

[0193] An input module for receiving the characteristic information of a subject with ASD, wherein the characteristic information includes: (i) the baseline expression level of a group of cytokines in the subject; (ii) the baseline behavioral performance of the subject; and (iii) the clinical information of the subject, including gender, age, and BMI, etc.;

[0194] A classification module configured to include a classifier, wherein the classifier can use the classifier to predict the response of the subject to bumetanide based on the characteristic information; and

[0195] An output module for outputting the prediction result.

[0196] In some embodiments, the prediction device may further include a training module configured to train the classifier using a training data set.

[0197] In some embodiments, the prediction device may further include a verification module configured to verify the classifier.

[0198] The present invention also provides a computer-readable medium, which includes computer-executable instructions recorded thereon for performing the following operations:

[0199] Receiving the characteristic information of a subject with ASD, wherein the characteristic information includes: (i) the baseline expression level of a group of cytokines of the subject; (ii) the baseline behavioral performance of the subject; and (iii) the clinical information of the subject, including gender, age, and BMI, etc.;

[0200] Predict the response of a subject to bumetanide using a classifier algorithm based on the said characteristic information.

[0201] In some embodiments, the operation further includes training the classifier using a training data set.

[0202] In some embodiments, the operation further includes validating the classifier.

[0203] In the said prediction device and computer-readable medium, the definitions related to the characteristic information and the classifier are the same as those in the previous section.

[0204] According to the present invention, based on the predicted response of the subject to bumetanide, it can be determined whether the subject will benefit from bumetanide treatment or whether the patient is a candidate for bumetanide treatment. For example, "responsive", "highly responsive", "better responsive", "positive response" or "responder" means that the subject may benefit from bumetanide treatment, and "non-responsive", "low responsive", "least responsive", "negative response" or "non-responder" means that the patient is less likely to benefit from bumetanide treatment. Thus, a treatment method can be selected for the patient according to the prediction result. For example, bumetanide can be administered to a subject who may benefit from bumetanide treatment, while bumetanide cannot be administered to a subject who cannot benefit from bumetanide treatment, or other treatments not including bumetanide can be administered to a subject who cannot benefit from bumetanide treatment.

[0205] In some embodiments, the present invention provides a method for treating ASD in a subject, the method comprising:

[0206] Predict the response of a subject with ASD to bumetanide treatment according to the characteristic information of the subject, wherein the characteristic information includes: (i) the baseline expression level of a group of cytokines in the subject; (ii) the baseline behavioral performance of the subject; and (iii) the clinical information of the subject, including gender, age and BMI, etc.;

[0207] Select a treatment method for the subject according to the prediction result.

[0208] In the method for treating ASD, the definition related to the said characteristic information and the implementation of the prediction are the same as those described in the previous section.

[0209] In the present invention, it should be understood that for treating a subject with ASD using bumetanide, bumetanide can be administered to the subject in an effective amount and in an appropriate manner, which can be determined by a skilled clinician.

[0210] For example, bumetanide can be administered parenterally or non-parenterally, such as orally, intravenously, intramuscularly, or by any other suitable route. Bumetanide can be formulated into dosage forms suitable for the above-mentioned routes of administration. For example, the dosage forms include dosage forms suitable for oral administration, such as tablets, capsules, caplets, pills, lozenges, powders, syrups, elixirs, suspensions, solutions, emulsions, sachets, and cachets; or dosage forms suitable for parenteral administration, such as sterile solutions, suspensions, and reconstituted powders.

[0211] For example, bumetanide can be administered to a subject at a total daily dose of about 0.5 to 10 mg, preferably 1 to 6 mg, more preferably 2 to 4 mg (divided into one, two or three doses). A dosage form containing 0.5, 1, 2 mg of bumetanide or a pharmaceutically acceptable salt thereof can be orally administered to a patient once, twice or three times a day. Single dose administration can enhance patient compliance, while multiple small doses can ensure constant serum levels.

[0212] The treatment period of bumetanide should be sufficient to demonstrate its therapeutic effect on the subject. The duration of treatment may be about 1 month to 6 months, such as 2 to 3 months, 2 to 4 months.

[0213] The present invention also provides a kit for predicting the response of an ASD patient to bumetanide. The kit may include reagents for measuring the expression level of a set of cytokines described in the present invention, such as antibodies that can specifically bind to the cytokines. The kit may also include instructions for the prediction method in the form of a label or a separate insert. For example, the instructions may tell how to obtain the characteristic information of the subject and how to predict the subject's response to bumetanide, for example as described in the present invention. In some embodiments, the kit may include one or more containers for containing reagents.

[0214] In some embodiments, the kit may further comprise bumetanide, and the kit may be used to treat ASD in a subject. The kit may further comprise one or more containers for containing bumetanide. The instructions may also inform how to treat the subject based on the predicted results, e.g., as described herein.

[0215] In the kit, the definitions related to feature information and the implementation of prediction are the same as described in the previous section.

[0216] Examples

[0217] Materials and methods

[0218] Participants

[0219] ASD participants were recruited from the Shanghai Xinhua ASD Registry at Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine in China, including participants from two previously registered clinical studies, namely CHICtr-OPC-16008336 and NCT03156153. According to the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5), patients were diagnosed with ASD. The diagnosis was confirmed by a total score of not less than 30 on the Autism Diagnostic Observation Schedule (ADOS) and the Childhood Autism Rating Scale (CARS). Exclusion criteria included impaired liver and kidney function; a history of sulfonamide drug allergy; abnormal electrocardiogram; genetic or chromosomal abnormalities; and neurological diseases (such as epilepsy, etc.). All patients underwent a comprehensive behavioral assessment and clinical sample collection. Between May 1, 2018 and April 30, 2019, a total of 90 ASD children aged 3 - 10 years received a three-month treatment with bumetanide, without behavioral intervention and without taking any psychoactive drugs simultaneously, and blood sampling and behavioral assessment were performed. Eleven of them were further excluded due to lack of 3-month follow-up data. Therefore, the current analysis used a subsample of 79 young children with ASD, whose blood samples were available before and after treatment. Blood samples were sent in two batches (discovery set: n = 37, December 4, 2019; validation set: n = 42, May 22, 2019) to measure the levels of 48 immunoreactive cytokines in plasma, and clinical symptoms were evaluated using the CARS, ADOS, and Social Responsiveness Scale (SRS). These 48 cytokines included basic FGF (bFGF), β-NGF, CTACK, Eotaxin, G-CSF, GM-CSF, GRO-α, HGF, IFN-α2, IFN-γ, IL-10, IL-12p40, IL-12p70, IL-13, IL-15, IL-16, IL-17, IL-18, IL-1α, IL-1β, IL-1Ra, IL-2, IL-2Rα, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IP-10, LIF, M-CSF, MCP-1, MCP-3, MIF, MIG, MIP-1α, MIP-1β, PDGF-BB, RANTES, SCF, SCGF-β, SDF-1α, TNF-α, TNF-β, TRAIL, and VEGF, of which 35 were used for subsequent operations (as shown in the figure below). The 35 cytokines and their clinical evaluations are shown in Table 1.

[0220] Table 1. Measurement of immune factors, clinical evaluations, and demographic parameters

[0221]

[0222] According to the protocol of previous studies 8,The bumetanide treatment consisted of taking two 0.5 mg tablets daily for three months, at 8:00 am and 4:00 pm respectively. The tablets were quite small, with a diameter of 8 mm and a thickness of 2 mm. Each time, the patient only needed to take half a tablet, which was not difficult for most patients. However, it was recommended that doctors, if necessary, grind the half-tablet into powder and then dissolve the powder in water for administration. Side effects that might occur were closely monitored during the treatment. Blood parameters (serum potassium and uric acid) were monitored through laboratory tests (Table 2), and symptoms (thirst, diuresis, nausea, vomiting, diarrhea, constipation, rash, palpitations, headache, dizziness, shortness of breath, and any other self-reported symptoms) were monitored through telephone interviews (Table 3), and reported to the research team by telephone 1 week and 1 month after the start of treatment and at the end of treatment. The cytokine levels of children with gastrointestinal problems were compared with those of children without gastrointestinal problems (Table 4). CAR and ADOS behavioral assessments and cytokine level measurements were performed at baseline before treatment and after 3 months of treatment. The SRS behavioral assessment was only used at baseline. This study was conducted in accordance with the provisions of the Declaration of Helsinki and the Good Clinical Practice guidelines and was approved by the Ethics Committee of Xinhua Hospital, Shanghai Jiao Tong University School of Medicine. Written informed consent was obtained from the parents or legal guardians of each participant before sample collection.

[0223] Table 2. Side effects measured by blood parameters reported during treatment*.

[0224]

[0225]

[0226] Table 2 (continued).

[0227]

[0228]

[0229] * Score = 0 indicates no symptoms; K score = 1 indicates hypokalemia (<3.5 mmol / L); U score = 1 indicates increased urine excretion (>417 μmol / L).

[0230] Before treatment, at baseline; 1 week, 1 week after bumetanide treatment; 1 month, 1 month after bumetanide treatment; 3 months, 3 months after bumetanide treatment.

[0231] Table 3. Side effects measured by symptoms reported during treatment*.

[0232]

[0233]

[0234] Table 3 (continued).

[0235]

[0236]

[0237] Scores of 0, 1, 2, and 3 represent asymptomatic, mild, moderate, and severe symptoms, respectively.

[0238] Table 4. Researchers compared cytokine levels between children with gastrointestinal diseases and children without gastrointestinal diseases.

[0239]

[0240]

[0241] 1 The data [i.e., mean (SD)] were first normalized and then corrected for batch effects.

[0242] 2 Mann-Whitney U test with multiple group comparisons.

[0243] 3 FDR adjustment was used for multiple tests.

[0244] Measurements

[0245] Clinical assessment

[0246] The CARS was used to diagnose and evaluate the severity of clinical symptoms in ASD patients. The CARS consists of 15 items scored from 1 to 4 on a 7-point scale; the higher the score, the higher the severity. The total score ranges from a minimum of 15 to a maximum of 60; scores below 30 indicate that an individual is in the non-autistic range, scores between 30 and 36.5 indicate mild to moderate autism, and scores between 37 and 60 indicate severe autism. We further divided these items into three subscales 28 : social impairment, negative affect, and distorted sensory responses. The ADOS, as a supplement to measure disease severity, includes total score items and 4 modules for evaluating social interaction, communication, play, and creative use of materials in individuals suspected of having ASD 29 . Through the observation of children's behaviors in the natural social environment, the SRS detected various mutual social behavior deficits ranging from absent to severe, focusing on the behaviors of children or adolescents between 4 and 18 years old. This is a questionnaire containing 65 questions filled out by teachers, parents, and / or other adult caregivers. A Likert four-point scale was used for scoring. Five subscales were also provided: social awareness (AWA), social cognition (COG), social communication (COM), social motivation (MOT), and autistic mannerisms (MANN)30 。

[0247] Cytokine levels

[0248] Peripheral blood was collected from each subject and centrifuged at 2300 rpm for 10 minutes to obtain plasma. The plasma was aliquoted and stored at -70 °C until cytokine analysis. Forty-eight cytokines, chemokines, and growth factors were measured using the Bio-Plex multiplex immunoassay (Bio- BIO-RAD Laboratories, Inc.). Before establishing our assay, we calibrated the Bio-Plex 200 system plate reader instrument according to the manufacturer's instructions. To prepare experimental samples, frozen plasma aliquots were thawed passively to room temperature and diluted four-fold in assay buffer (15 μL sample + 45 μL sample diluent HB). After preparing the capture bead mixture and standards, the immunoassay was performed on a 96-well plate. The experimental procedure was carried out according to the instructions. Data acquisition was set to at least 50 bead counts per well per analyte. Using a standard curve derived from known reference cytokine concentrations provided by the manufacturer, the cytokine concentrations of unknown samples were processed and presented by Bio-plex Manager software. The final concentration was calculated using a five-parameter model and expressed in pg / ml. The sensitivity of this assay could detect cytokine concentrations in the following ranges: IFN-α 23.6 - 3992.4 pg / ml; IFN-γ 0.9 - 14556.8 pg / ml; IL-1β 0.3 - 5375.0 pg / ml; IL-4 0.2 - 3455.0 pg / ml; IL-6 0.4 - 5961.8 pg / ml; IL-8 0.2 - 15570.4 pg / ml; MIG 5.7 - 30955.2 pg-ml; TNF-α 3.3 - 52256.0 pg / ml, etc. All subsequent analyses excluded concentrations below the limit of detection (LOD) of this method.

[0249] Statistical analysis

[0250] Data preprocessing

[0251] After excluding cytokines with values below the limit of detection, our analysis included 35 cytokines. For two batches of data including baseline and changes in cytokine levels (the difference between baseline and subsequent data), min-max normalization and logarithmic transformation were performed separately. To adjust for batch effects, we applied the empirical Bayes method to the cytokine level baseline using the "ComBat" parametric algorithm provided in the R package "sva" 31 and used principal component analysis (PCA) to visualize non-biological variations due to batch effects and repeated it to confirm the adjustment ( Figure 4)。Before and after treatment, we compared the demographic parameters (i.e., sex ratio, age, body mass index [BMI]) and symptom severity (i.e., ADOS and CARS) between the two groups of data using t-tests, Mann-Whitney U tests, or Pearson chi-square tests (where applicable).

[0252] Multivariate association analysis characterizes immune behavior covariation

[0253] First, the pairwise correlations between the CARS_total score and the levels of 35 cytokines were evaluated by Spearman rank correlation. The correlations between the changes in the CARS_total score and the changes in the levels of 35 cytokines after treatment were also tested. The false discovery rate (FDR) was used to correct for multiple comparisons.

[0254] n×p 1 Second, to reveal the multivariate associations between behavioral assessments and cytokine levels, we employed sparse canonical correlation analysis (sCCA) provided in the R package "sRDA" (version 1.0.0) 32 n×p 2 Canonical correlation analysis (CCA) is a classical method for determining the relationship between two sets of variables. Given two data sets X 1 and X 2 , with dimensions and respectively, from n observations, CCA seeks pairs of linear combinations (i.e., canonical variables), one from the variables in X 1 and the other from the variables in X 2 that have the maximum correlation with each other. However, the contributions of some variable pairs to the canonical variables may be negligible but non-zero. sCCA was developed to address this issue. sCCA applies an L 1 penalty to the canonical weights, forcing some of these weights to take zero values. Mathematically,

[0255]

[0256] Here, assume c 1 and c 2 belong to the boundaries and, where p 1 and p 2 are the number of features in X1 and X2 respectively. We refer to w 1 and w 2 as the canonical weights, and X 1 w 1 and X 2 w 2It is called the canonical score. Therefore, this algorithm can identify a linear combination of the three CARS subscales (i.e., the behavioral component), which is significantly correlated with another linear combination of several cytokine levels (i.e., the cytokine component). At the same time, the sparsity of this algorithm ensures that only the key cytokines driving the behavioral associations are selected in the immune component.

[0257] In the discovery set, we explored the sCCA between the CARS subscales and cytokine levels using baseline data or the changes between baseline and follow-up. The significance of the identified canonical correlations was evaluated by 5000 permutations. 32 Only the significant canonical components were retained. Sensitivity analysis was performed using the data in the validation set by re-evaluating the canonical correlations after controlling for potential confounding factors such as age, sex, BMI, and the canonical variables at baseline. Using the cytokine component score (x-axis) and behavioral component score (y-axis) established by sCCA, each patient can be mapped onto a two-dimensional plane, called the immune-behavioral covariation plane, which characterizes the immune-behavioral covariation in the response pattern of autistic children to bumetanide treatment.

[0258] Cluster analysis identifies immune-behavioral groups

[0259] We applied the unsupervised clustering algorithm k-means to identify patient clusters based on immune-behavioral covariation. Patients in each cluster (i.e., the immune-behavioral group within ASD) have similar canonical scores, indicating that they have similar response patterns to bumetanide in terms of the immune system and clinical behavior. The cluster structure was first identified using the discovery set and then validated using the validation set. The optimal number of clusters was selected based on the elbow of the scree plot (maximum change) using the Hubert statistic implemented in the R package "NBclust". 33 。

[0260] To demonstrate the different response patterns of the immune-behavioral groups to bumetanide, we performed one-sample t-tests on the post-treatment changes in the CARS subscales and cytokines selected by the above sCCA algorithm. We also used the Kruskal-Wallis rank sum test and FDR correction for multiple comparisons to compare these changes between the above-identified immune-behavioral groups.

[0261] Using baseline information to predict the treatment response to bumetanide

[0262] To predict the response of autistic patients to bumetanide, we used the pre-treatment baseline information to train a classifier for the above-defined immune-behavioral groups. The baseline information included 35 cytokine levels, 3 clinical assessments (CARS, SRS, ADOS), and 2 demographic parameters (gender and age). The classifiers included oblique random forest (ORF), partial least squares (PLS), sparse linear discriminant analysis (sLDA), neural network (NN), and support vector machine (SVM), which were implemented by the R package "caret" and had feature selection and oversampling capabilities. 34 The model was first trained using the discovery set and then its performance was compared using the validation set. A 95% confidence interval for the difference in the area under the curve of a pair of models was constructed by the 100-bootstrap method.

[0263] First, we tested whether the data including baseline cytokine levels could improve the prediction of immune-behavior-defined responders. The average performance (i.e., average area under the curve) of the five classifiers was reported and compared. General

[0264] AUC

[0265] Second, we tested whether behavior-defined responders were more difficult to predict at baseline compared to immune-behavior-defined responders. In a previous clinical trial of bumetanide for ASD treatment, 5-7 the proportion of patients who responded positively to the treatment was between 30% and 40%. Therefore, we divided the ASD patients into two groups based on ΔCARS_total (= baseline CARS_total – follow-up CARS_total), with cut-off values of 2.5 (n = 22, 27.85% of patients had ΔCARS_total > 2.5) or 2 (n = 32, 40.51% of patients had ΔCARS_total > 2).

[0266] Results

[0267] Participants

[0268] This study used data from two groups of children with ASD (n = 79). The average age of dataset 1 (n = 37) was 47 months (±17.35 months), and 18.92% were girls; the average age of dataset 2 (n = 42) was 54 months (±20.19 months), and 23.81% were girls. There were no significant differences in clinical characteristics or cytokine levels between the two datasets (Table 5; Table 6).

[0269] Table 5. Demographic and clinical (mean (SD)) characteristics of the two datasets

[0270]

[0271] ​

[0272] 1 The normal features are represented by the T - test statistic, the abnormal features are represented by the Mann - Whitney U - test, and the gender is represented by the chi - square test.

[0273] 2 The sample sizes of ADO data in the discovery set and the validation set are 36 and 41 respectively.

[0274] 3 The sample sizes of SRS data in the discovery set and the validation set are 21 and 39 respectively.

[0275] BMI, body mass index; CARS, Childhood Autism Rating Scale; ADOS, Autism Diagnostic Observation Schedule; SRS, Social Responsiveness Scale; CARS_total, total score of CARS; CARS_S, score of CARS in the social impairment domain; CARS_N, score of CARS in the negative affect domain; CARS_D, score of CARS in the distorted sensory responses domain; ADOS_S, score of ADOS in social interaction; ADOS_C, score of ADOS in communication; ADOS_P, score of ADOS in play; ADOS_I, score of ADOS in the creative use of materials; SRS_AWA, score of SRS in social awareness; SRS_COG, score of SRS in social cognition; SRS_COM, score of SRS in social communication; SRS_MOT, score of SRS in social motivation; SRS_MANN, score of SRS in autistic mannerisms; SRS_total, total score of SRS;

[0276] Table 6. Baseline levels of cytokines in two groups.

[0277]

[0278]

[0279] 1 The data [i.e., mean (SD)] were first standardized and then corrected for batch effects.

[0280] 2 Mann - Whitney U - test.

[0281] 3 FDR adjustment was used for multiple tests.

[0282] Changes after bumetanide administration

[0283] All patients received bumetanide treatment for 3 months, and the total CARS score decreased after treatment (effect size Cohen’s d = 1.26, t 78 = 11.21, p < 0.001). The treatment effects showed no difference between the two groups of data (ΔCARS_total: mean (±SD): 1.54 (±1.40) vs. 1.90 (±1.34)). Consistent with previous studies on low-dose bumetanide treatment for ASD, side effects were rarely reported (Table 2-3). At baseline, there were no significant differences in cytokine levels between children with and without gastrointestinal problems (Table 4). After treatment, many cytokine levels changed significantly (Table 7). No significant pairwise associations were found in the discovery set, and it was also not possible to verify significant pairwise associations among the four groups of variables: baseline CARS total score, baseline cytokine levels, change in CARS total score, and change in cytokine levels using the validation set( Figure 5 ).

[0284] Table 7. Levels of cytokine changes in the two groups.

[0285]

[0286]

[0287] 1 The degrees of freedom of the one-sample t-test statistic were 36.

[0288] 2 The degrees of freedom of the one-sample t-test statistic were 41.

[0289] 3 Mann-Whitney U test.

[0290] 4 FDR adjustment was used for multiple tests.

[0291] Covariation between symptom improvement and cytokine changes

[0292] Using the discovery set, we found a canonical correlation (r = 0.459; by permutation p < 0.001; Figure 1 A) between two canonical components (i.e., cytokine component and behavioral component) by sCCA. The cytokine component was a combination of changes in the levels of 3 cytokines, including MIG, IFN-α2, and IFN-γ. The behavioral component was a combination of changes in 3 subscales of the CARS score, including social impairment score, negative mood score, and distorted sensory response score. Applying these canonical weights to an independent dataset (i.e., the validation set), we confirmed the correlation between the cytokine component and the behavioral component (r = 0.316; p = 0.012 (by permutation); Figure 1 B).

[0293] Sensitivity analysis

[0294] Using the validation set, we found that the correlations between the identified canonical components remained significant (Table 8). At baseline, we also found that the cytokine canonical scores were correlated with other clinical assessments, including SRS_total (r = -0.269; t 63 = -2.21, p = 0.031), ADOS_S (r = 0.296; t 76 = 2.70, p = 0.009), ADOS_P (r = -0.244; t 76 = -2.19, p = 0.032), and ADOS_I (r = -0.251; t 76 = -2.26, p = 0.027).

[0295] Table 8. Correlations between the identified canonical components after controlling for baseline variables.

[0296] Sensitivity analysis was performed by re - evaluating this correlation after controlling for baseline potential confounders.

[0297]

[0298]

[0299] 1 The degrees of freedom of the student t - test statistic were 37.

[0300] * The p - value was less than 0.05.

[0301] CARS_S, CARS score in the social impairment domain; CARS_N, CARS score in the negative affect domain; CARS_D, CARS score in the distorted sensory response domain; IFN - γ, interferon - γ; IFN - α2, interferon - α2; MIG, monokine induced by interferon - γ; BMI, body mass index.

[0302] Three different response patterns revealed by immune - behavior covariation

[0303] Using the CARS and cytokine component scores, we found that the patients were divided into 3 clusters ( Figure 2 A - B; Figure 6 ). There was no significant difference in the distribution of patients among these 3 clusters between the discovery set and the validation set. In addition, there were no significant differences in the CARS score baseline and cytokine levels among these 3 groups. Comparing these three groups of patients in two sets ( Figure 2 C - D; Table 9; Figure 7) We found that the best response group (n = 17) had the largest reduction in the total CARS score (ΔCARS_total: 3.32 (±1.47), Hedge's g = 2.16, p < 0.001), with the most prominent reduction in the social impairment score (ΔCARS_S: 2.15 (±1.13), g = 1.81, p < 0.001) and elevated cytokine levels (ΔMIG: g = 0.72, p = 0.012; ΔIFN-α2: g = 0.84, p = 0.006). The worst response group (n = 18) had the smallest reduction in the total CARS score (ΔCARS_total: 1.03 (±0.96), g = 1.02, p < 0.001), with the most significant reduction in the negative mood score (ΔCARS_N: 0.69 (±0.55), g = 1.21, p < 0.001), and the total CARS score of the worst response group also decreased (ΔMIG: g = -1.07, p < 0.001; ΔIFN-γ: g = -1.32, p < 0.001). The total CARS score and each subscale in the moderate response group decreased significantly, but the effect was small, and the IFN-γ level decreased and the IFN-α2 level increased in this group (Table 9).

[0304]

[0305] Baseline cytokine levels help identify treatment responders

[0306] Training five different types of classifiers using the discovery set and testing the performance using the validation set, we found that the prediction accuracy including baseline cytokine levels significantly improved for the best response group (with / without baseline cytokine levels = 0.768 / 0.646; improvement, 95% confidence interval (CI): (0.103, 0.130); Figure 3 A - B) and the worst response group (with / without baseline cytokine levels = 0.698 / 0.618; improvement, 95% CI: (0.064, 0.097); Figure 3 C - D). After incorporating cytokine levels into the model, all five models showed better performance (Table 10).

[0307] Among them,

[0308] 35 cytokines introduced by the model: IL1beta, IL6, IL8, IFNgamma, TNFalpha, MCP1 / MCAF, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGF-BB, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, SCGFβ;

[0309] Other variables introduced by the model: age, gender, CARS_total, CARS_S, CARS_N, CARS_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, SRS_MANN, SRS_total;

[0310] All variables were at baseline levels.

[0311] The output is whether the subject belongs to the best response group defined above.

[0312] In Table 10, "including" means including 35 cytokines and the other variables mentioned above, and "not including" means including the other variables mentioned above and not including 35 cytokines.

[0313] We also trained 5 different types of classifiers using the discovery set, tested the performance using the validation set, and used the following variables to predict behavior-defined responders and immune-behavior-defined responders.

[0314] 35 cytokines introduced by the model: IL1β, IL6, IL8, IFNγ, TNFα, MCP1 / MCAF, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGF-BB, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, SCGFβ; Other variables introduced by the model: age, gender, CARS_total, CARS_S, CARS_N, CARS_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, SRS_MANN, SRS_total; All variables were at baseline levels.

[0315] To define behaviorally defined responders, we divided the ASD patients into two groups based on ΔCARS_total (= baseline CARS_total – follow-up CARS_total), with cut-off values of 2.5 (n = 22, 27.85% of patients with ΔCARS_total > 2.5) or 2 (n = 32, 40.51% of patients with ΔCARS_total > 2) points. These two groups were defined as behaviorally defined responders.

[0316] The definition of immunologically behaviorally defined responders is as described above in "Three different response patterns revealed by immunological behavior covariation". Specifically, to define immunologically behaviorally defined responders, we applied the unsupervised clustering algorithm k-means to identify patient clusters based on immunological behavior covariation. Patients within each cluster (i.e., immunological behavior groups within ASD) had similar canonical scores, indicating that they had similar response patterns to bumetanide in terms of the immune system and clinical behavior. The cluster structure was first identified using the discovery set and then validated using the validation set. The optimal number of clusters was selected based on the elbow (maximum change) of the scree plot using the Hubert statistic implemented in the R package "NBclust" 33 。

[0317] To demonstrate the different response patterns of the immunological behavior groups to bumetanide, we performed one-sample t-tests on the post-treatment changes in the CARS subscale and cytokines selected by the above sCCA algorithm. We also used the Kruskal-Wallis rank sum test and FDR correction for multiple comparisons to compare these changes between the immunological behavior groups determined above. Responders defined by immunological behavior were divided into the best response group, moderate response group, and worst response group.

[0318] We found that these five models could accurately predict the behaviorally defined responder population with high precision (Table 11; Figure 8 A). Similar results were also found when using a threshold of 2.5 for the change in CARS score to behaviorally define responders ( Figure 8 B).

[0319] In addition, we found that immunologically behaviorally defined responders had higher prediction accuracy at baseline compared to behaviorally defined responders.

[0320] Table 10. Comparison of AUC between models with and without cytokines to predict the best response group and worst response group defined by immunological behavior.

[0321]

[0322] ORF - oblique random forest; SVM - support vector machine; PLS - partial least squares; sLDA - sparse linear discriminant analysis; NN - neural network

[0323] Table 11. Comparison of AUC between the models for the responder groups defined by predicted immune behavior and those defined by behavioral response

[0324]

[0325] 1 The model was constructed with 100 bootstrap samples.

[0326] AUC, area under the ROC curve; CI, confidence interval; SD, standard deviation; ORF, oblique random forest; PLS, partial least squares; SVM, support vector machine; sLDA, sparse linear discriminant analysis; NN, neural network; ΔAUC, difference in AUC between the models with and without each cytokine.

[0327] Using the five models established above, we illustrate the effectiveness of these models through some examples. For example:

[0328] We used the support vector machine (SVM) and found that a prediction model with three cytokines (i.e., IL16, GROalpha, and IL7) and seven other variables (i.e., age, gender, ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D) achieved a high accuracy of 0.833 in predicting the best responder group (area under the curve, AUC = 74.0%). ( Figure 9 )

[0329] We used partial least squares (PLS) and found that a prediction model with three cytokines (i.e., IL16, GROalpha, and TNFbeta) and six other variables (i.e., age, gender, ADOS_S, ADOS_C, ADOS_P, and CARS_total) achieved a high accuracy of 0.833 in predicting the best responder group (area under the curve, AUC = 74.0%). ( Figure 9 )

[0330] We used the neural network (NN) and found that a prediction model with five cytokines (i.e., IL16, GROalpha, IL7, CTACK, and TNFbeta) and nine other variables (i.e., age, gender, ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT) achieved a high accuracy of 0.810 in predicting the best responder group (area under the curve, AUC = 74.0%). ( Figure 9 )

[0331] Using sparse linear discriminant analysis (sLDA), we found that a prediction model with six cytokines (i.e., IL16, GROalpha, IL7, LIF, MIF, and TNFbeta) and eight other variables (i.e., age, gender, ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total) achieved a high accuracy of 0.786 (area under the curve, AUC = 76.2%) in predicting the best response group. Figure 9 )

[0332] Using oblique random forest (ORF), we found that a prediction model with three cytokines (i.e., IL16, GROalpha, and TNFbeta) and six other variables (i.e., age, gender, ADOS_S, ADOS_C, ADOS_P, and CARS_total) achieved a high accuracy of 0.810 (area under the curve, AUC = 68.2%) in predicting the best response group.

[0333] ( Figure 9 )

[0334] Discussion

[0335] In this study, we observed a significant improvement in the clinical symptoms of autistic children treated with bumetanide, and this improvement was related to the change patterns of three cytokine levels, namely IFN-γ, MIG, and IFN-α2 (r = 0.459 in the discovery set and r = 0.316 in the validation set). Compared with using behavioral assessment alone, these baseline cytokine levels could improve the prediction of responders to bumetanide, and the best predictors achieved an AUC of 0.83 in the independent test dataset (Table 10). The significance of these findings may be twofold: 1) A large part of the heterogeneity in the clinical effects of bumetanide treatment for autistic children is related to differences in the patient immune system; 2) The compositional scores of cytokines have the potential to construct a blood signature for predicting and monitoring the therapeutic effects of bumetanide on autistic children.

[0336] We found immune-behavior covariation, highlighting the role of the immune system in the clinical effects of bumetanide on young children with ASD. IFN-γ is a T helper cell 1 (Th1) cytokine with pro-inflammatory effects and was selected by the sCCA algorithm as one of the three cytokines forming a canonical score related to the improvement of CARS. Compared with the control group, the content of IFN-γ in the brain tissue 22 , cerebrospinal fluid (CSF) 23 , plasma 35 and peripheral blood mononuclear cells (PBMC) 36 of ASD patients was higher, while the content of IFN-γ in the neonatal dried blood spot samples (n-DBSS) of ASD children was lower37 Increasing evidence suggests that IFN-γ can inhibit chloride secretion 38 and downregulate NKCC1 expression 16,38 and Na+-K+-ATPase expression 16 , which is associated with GABAergic dysfunction in ASD 10,39 . Indirectly, animal studies have also shown that high concentrations of IFN-γ stimulation can increase the expression of IL-1β 40 , an inflammatory cytokine that can affect the expression of chloride transporters and delay the developmental switch of GABA signaling 17 . Therefore, the immune system can interact with the mechanism of action of bumetanide to restore GABA function in ASD.

[0337] The association between cytokines and symptoms was confirmed in the changes after bumetanide treatment, but not found before treatment, indicating that bumetanide may interact with cytokines, and the changes in cytokines contribute to the therapeutic effect of bumetanide. Animal studies have shown that the brain efflux rate of bumetanide is very fast, but many clinical trials have shown that the drug has significant therapeutic effects on neuropsychiatric diseases including autism, epilepsy and depression 41,42 . These findings may indicate that bumetanide, as a neuromodulator for these neuropsychiatric diseases, may have systemic effects. Considering its molecular structure, bumetanide was recently identified by small molecule in vitro screening as having anti-inflammatory drug effects through interleukin inhibition 43 . This anti-inflammatory activity of bumetanide may alter the blood levels of cytokines outside the blood-brain barrier (BBB). In fact, it has been reported that in RAW264.7 cells and mice with lung injury, after direct pulmonary administration, bumetanide can reduce the production of lipopolysaccharide-induced pro-inflammatory cytokines 44 . These inflammatory signal messengers can cross the BBB 45 and affect neuronal chloride homeostasis by altering KCC2 expression 18 and other ways. The feasibility of reducing inflammation to enhance KCC2 expression was discussed in a review in 2020 17 . In fact, we found that the cytokine component score increased the most in the best response group (Hedge’s g = 2.16, reduction in the total CARS score). In contrast, the cytokine component score decreased the most in the worst response group (g = 1.02). Collectively, these findings suggest that bumetanide may be a drug that inhibits NKCC1 and enhances KCC2 by interacting with cytokines inside and / or outside the brain.

[0338] Our research results suggest that the established cytokine signature has the potential to construct a blood-based signature for predicting and monitoring bumetanide treatment in children with autism. Accurately identifying patients who are likely to respond positively to bumetanide can facilitate precision medicine for ASD. Our prediction model based on pre-treatment cytokine levels could provide a potentially new tool for precision medicine in ASD. Given the genetic heterogeneity of ASD, accurately identifying subsets of ASD patients who are likely to respond positively to its pharmacological treatment has important clinical value. 46 Multiple factors, including genetic and environmental factors, may contribute to the heterogeneity of ASD and its response to treatment. 12,19,47 For example, prenatal insults (including maternal infections and subsequent immune activation during pregnancy) may increase the risk of autism in children. Exposure to higher levels of air pollution during pregnancy has also been associated with abnormal mitochondrial metabolism in childhood, which may also increase the risk of ASD. However, our study shows that using cytokine levels can improve the prediction of the response to bumetanide treatment for ASD in three ways: 1) Immune-behavior covariation enables the identification of a more homogeneous subgroup of ASD in terms of response to bumetanide. We demonstrated consistent evidence using five models and an independent test dataset that the responder group identified by immune-behavior covariation can be better predicted by baseline information compared to the responder group defined only by CARS. 2) By combining cytokine levels with pre-treatment CARS, ADOS, and SRS clinical assessments, we achieved a higher accuracy of 84.3% in identifying ASD children who are likely to respond positively to bumetanide treatment. 3) Blood cytokine levels are more easily obtainable in clinical practice.

[0339] Our study has several limitations. Although we had two independent datasets to validate our findings, the sample size was limited. It has been previously reported that there are gender differences in cytokine-symptom associations, but due to the small number of girls with ASD in our sample, we were unable to test for such gender differences. 48 Therefore, future studies with a multi-center, prospective, and larger sample size are needed to confirm the current findings. Secondly, the hypothesized molecular mechanisms underlying the efficacy of bumetanide treatment for ASD need to be causally confirmed in animal studies.

[0340] In summary, we have established the association between changes in cytokine levels and symptom improvement in autistic children following bumetanide treatment and found that the therapeutic effect of bumetanide can be better characterized by immune-behavioral covariation. This finding may provide new clinically important evidence to support the hypothesis that the immune response may interact with the mechanism of bumetanide to restore GABAergic function in ASD. This finding may also help to identify enriched samples of children with ASD to participate in research on novel drug therapies with a mode of action similar to that of bumetanide but potentially higher efficacy and fewer side effects.

[0341] Data and code availability

[0342] The data of this study are available to the corresponding author upon reasonable and ethical request. The R code of this study can be found at https: / / github.com / qluo2018 / ImmunoASD4BTN.

[0343] References:

[0344] 1. Lord C, Elsabbagh M, Baird G, Veenstra-Vanderweele J. Autism spectrum disorder. Lancet 2018; 392(10146): 508 - 520.

[0345] 2. Hirvikoski T, Mittendorfer-Rutz E, Boman M, Larsson H, Lichtenstein P, S. Premature mortality in autism spectrum disorder. Br J Psychiatry 2016; 208(3): 232 - 238.

[0346] 3. Phillips AG. Placing old wine into new bottles: Successful repurposing of bumetanide for treatment of autism spectrum disorder. Science Bulletin 2021.

[0347] 4. James BJ, Gales MA, Gales BJ. Bumetanide for Autism Spectrum Disorder in Children: A Review of Randomized Controlled Trials. Ann Pharmacother 2019;53(5):537 - 544.

[0348] 5. Zhang L, Huang C - C, Dai Y, Luo Q, Ji Y, Wang K et al. Symptom improvement in children with autism spectrum disorder following bumetanide administration is associated with decreased GABA / glutamate ratios. Transl Psychiatry 2020;10(1):9.

[0349] 6. Lemonnier E, Degrez C, Phelep M, Tyzio R, Josse F, Grandgeorge M et al. A randomised controlled trial of bumetanide in the treatment of autism in children. Transl Psychiatry 2012;2:e202.

[0350] 7. Lemonnier E, Villeneuve N, Sonie S, Serret S, Rosier A, Roue M et al. Effects of bumetanide on neurobehavioral function in children and adolescents with autism spectrum disorders. Transl Psychiatry 2017;7(3):e1056.

[0351] 8. Dai Y, Zhang L, Yu J, Zhou X, He H, Ji Y et al. Improved symptoms following bumetanide treatment in children aged 3 to 6 years with autism spectrum disorder: a randomized, double-blind, placebo-controlled trial. Science Bulletin 2021.

[0352] 9. Sprengers JJ, van Andel DM, Zuithoff NPA, Keijzer-Veen MG, Schulp AJA, Scheepers FE et al. Bumetanide for Core Symptoms of Autism Spectrum Disorder (BAMBI): A Single Center, Double-Blinded, Participant-Randomized, Placebo-Controlled, Phase-2 Superiority Trial. Journal of the American Academy of Child & Adolescent Psychiatry 2021; 60(7):865-876. 10. Coghlan S, Horder J, Inkster B, Mendez MA, Murphy DG, Nutt DJ. GABA system dysfunction in autism and related disorders: from synapse to symptoms. Neuroscience & Biobehavioral Reviews 2012; 36(9):2044-2055.

[0353] 11. Chao H-T, Chen H, Samaco RC, Xue M, Chahrour M, Yoo J et al. Dysfunction in GABA signalling mediates autism-like stereotypies and Rett syndrome phenotypes. Nature 2010; 468(7321):263-269.

[0354] 12. Tyzio R, Nardou R, Ferrari DC, Tsintsadze T, Shahrokhi A, Eftekhari S et al. Oxytocin-mediated GABA inhibition during delivery attenuates autism pathogenesis in rodent offspring. Science 2014;343(6171):675-679.

[0355] 13. Lemonnier E, Ben-Ari Y. The diuretic bumetanide decreases autistic behaviour in five infants treated during 3 months with no side effects. Acta Paediatr 2010;99(12):1885-1888.

[0356] 14. Du L, Shan L, Wang B, Li H, Xu Z, Staal WG et al. A Pilot Study on the Combination of Applied Behavior Analysis and Bumetanide Treatment for Children with Autism. J Child Adolesc Psychopharmacol 2015;25(7):585-588.

[0357] 15. Jones KA, Thomsen C. The role of the innate immune system in psychiatric disorders. Mol Cell Neurosci 2013;53:52-62.

[0358] 16. Bertelsen LS, Eckmann L, Barrett KE. Prolonged interferon-gamma exposure decreases ion transport, NKCC1, and Na+-K+-ATPase expression in human intestinal xenografts in vivo. Am J Physiol Gastrointest Liver Physiol 2004;286(1):G157-G165.

[0359] 17. Pozzi D, Rasile M, Corradini I, Matteoli M. Environmental regulation of the chloride transporter KCC2: switching inflammation off to switch the GABA on? Transl Psychiatry 2020;10(1):349.

[0360] 18. Corradini I, Focchi E, Rasile M, Morini R, Desiato G, Tomasoni R et al. Maternal Immune Activation Delays Excitatory-to-Inhibitory Gamma-Aminobutyric Acid Switch in Offspring. Biol Psychiatry 2018;83(8):680-691.

[0361] 19. Meltzer A, Van de Water J. The Role of the Immune System in Autism Spectrum Disorder. Neuropsychopharmacology 2017;42(1):284-298.

[0362] 20. Saghazadeh A, Ataeinia B, Keynejad K, Abdolalizadeh A, Hirbod-Mobarakeh A, Rezaei N. Anti-inflammatory cytokines in autism spectrum disorders: A systematic review and meta-analysis. Cytokine 2019;123:154740.

[0363] 21. Kordulewska NK, Kostyra E, Piskorz-Ogórek K, Moszyńska M, A, Fiedorowicz E et al. Serum cytokine levels in children with spectrum autism disorder: Differences in pro- and anti-inflammatory balance. J Neuroimmunol 2019;337:577066.

[0364] 22. Li X, Chauhan A, Sheikh AM, Patil S, Chauhan V, Li X-M et al. Elevated immune response in the brain of autistic patients. J Neuroimmunol 2009;207(1-2):111-116.

[0365] 23. Vargas DL, Nascimbene C, Krishnan C, Zimmerman AW, Pardo CA. Neuroglial activation and neuroinflammation in the brain of patients with autism. Ann Neurol 2005;57(1):67-81.

[0366] 24. Ashwood P, Krakowiak P, Hertz-Picciotto I, Hansen R, Pessah IN, Van deWater J. Associations of impaired behaviors with elevated plasma chemokines in autism spectrum disorders. J Neuroimmunol 2011;232(1-2):196-199.

[0367] 25. Ashwood P, Enstrom A, Krakowiak P, Hertz-Picciotto I, Hansen RL, Croen LA et al. Decreased transforming growth factor beta1 in autism: a potential link between immune dysregulation and impairment in clinical behavioral outcomes. J Neuroimmunol 2008;204(1-2):149-153.

[0368] 26. Ashwood P, Krakowiak P, Hertz-Picciotto I, Hansen R, Pessah I, Van deWater J. Elevated plasma cytokines in autism spectrum disorders provide evidence of immune dysfunction and are associated with impaired behavioral outcome. Brain Behav Immun 2011;25(1):40-45.

[0369] 27. Masi A, Glozier N, Dale R, Guastella AJ. The Immune System, Cytokines, and Biomarkers in Autism Spectrum Disorder. Neurosci Bull 2017;33(2):194-204.

[0370] 28. DiLalla DL, Rogers SJ. Domains of the childhood autism rating scale: Relevance for diagnosis and treatment. Journal of Autism and Developmental Disorders 1994; 24(2): 115 - 128.

[0371] 29. Lord C, Risi S, Lambrecht L, Cook EH, Leventhal BL, DiLavore PC et al. The autism diagnostic observation schedule - generic: a standard measure of social and communication deficits associated with the spectrum of autism. Journal of autism and developmental disorders 2000; 30(3): 205 - 223.

[0372] 30. Constantino JN, Gruber CP. Social responsiveness scale: SRS - 2. Western Psychological Services Torrance, CA 2012.

[0373] 31. Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 2007; 8(1): 118 - 127.

[0374] 32. Csala A, Voorbraak FPJM, Zwinderman AH, Hof MH. Sparse redundancy analysis of high - dimensional genetic and genomic data. Bioinformatics 2017; 33(20): 3228 - 3234.

[0375] 33. Charrad M, Ghazzali N, Boiteau V, Niknafs A. NbClust: An R Package for Determining the Relevant Number of Clusters in a Data Set. Journal of Statistical Software 2014; 61: 1 - 36.

[0376] 34. Kuhn M. Building Predictive Models in R Using the caret Package. Journal of Statistical Software; Vol 1, Issue 5 (2008) 2008.

[0377] 35. Saghazadeh A, Ataeinia B, Keynejad K, Abdolalizadeh A, Hirbod - Mobarakeh A, Rezaei N. A meta - analysis of pro - inflammatory cytokines in autism spectrum disorders: Effects of age, gender, and latitude. J Psychiatr Res 2019; 115.

[0378] 36. Molloy CA, Morrow AL, Meinzen - Derr J, Schleifer K, Dienger K, Manning - Courtney P et al. Elevated cytokine levels in children with autism spectrum disorder. J Neuroimmunol 2006; 172(1 - 2): 198 - 205.

[0379] 37. Abdallah MW, Larsen N, Mortensen EL, Atladóttir - Pedersen B, Bonefeld - EC et al. Neonatal levels of cytokines and risk of autism spectrum disorders: an exploratory register-based historic birth cohort study utilizing the Danish Newborn Screening Biobank. J Neuroimmunol 2012;252(1-2):75-82.

[0380] 38. Resta-Lenert S, Barrett KE. Probiotics and commensals reverse TNF-alpha- and IFN-gamma-induced dysfunction in human intestinal epithelial cells. Gastroenterology 2006;130(3):731-746.

[0381] 39. Rivera C, Voipio J, Payne JA, Ruusuvuori E, Lahtinen H, Lamsa K et al. The K+ / Cl- co-transporter KCC2 renders GABA hyperpolarizing during neuronal maturation. Nature 1999;397(6716):251-255.

[0382] 40. Schmidt J, Barthel K, Wrede A, Salajegheh M, M, Dalakas MC. Interrelation of inflammation and APP in sIBM: IL-1 beta induces accumulation of beta-amyloid in skeletal muscle. Brain 2008;131(Pt 5):1228-1240.

[0383] 41. Kharod SC, Kang SK, Kadam SD. Off-Label Use of Bumetanide for Brain Disorders: An Overview. Front Neurosci 2019;13:310.

[0384] 42. Hu D, Yu Z-L, Zhang Y, Han Y, Zhang W, Lu L, et al. Bumetanide treatment during early development rescues maternal separation-induced susceptibility to stress. Scientific Reports 2017; 7(1): 11878.

[0385] 43. Wang Z, Wang Y, Vilekar P, Yang S-P, Gupta M, Oh MI, et al. Small molecule therapeutics for COVID-19: repurposing of inhaled furosemide. PeerJ 2020; 8: e9533-e9533.

[0386] 44. Hung C-M, Peng C-K, Wu C-P, Huang K-L. Bumetanide attenuates acute lung injury by suppressing macrophage activation. Biochemical Pharmacology 2018; 156: 60-67.

[0387] 45. Banks WA, Kastin AJ, Broadwell RD. Passage of cytokines across the blood-brain barrier. Neuroimmunomodulation 1995; 2(4): 241-248.

[0388] 46. Loth E, Murphy DG, Spooren W. Defining Precision Medicine Approaches to Autism Spectrum Disorders: Concepts and Challenges. Front Psychiatry 2016; 7: 188-188.

[0389] 47. Frye RE, Cakir J, Rose S, Delhey L, Bennuri SC, Tippett M et al. Prenatal air pollution influences neurodevelopment and behavior in autism spectrum disorder by modulating mitochondrial physiology. Mol Psychiatry 2021; 26(5): 1561 - 1577.

[0390] 48. Masi A, Breen EJ, Alvares GA, Glozier N, Hickie IB, Hunt A et al. Cytokine levels and associations with symptom severity in male and female children with autism spectrum disorder. Mol Autism 2017; 8: 63 - 63.

Claims

1. A method for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, the method comprises: obtaining characteristic information of the subject, wherein the characteristic information includes: (i) the baseline expression levels of a group of cytokines in the subject; (ii) the baseline behavioral performance of the subject; and (iii) the clinical information of the subject, including gender and age; predicting the response of the subject to bumetanide based on the characteristic information.

2. The method according to claim 1, wherein the group of cytokines comprises three or more cytokines selected from the group consisting of IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9 and SCGFβ.

3. The method according to claim 2, wherein the group of cytokines comprises a group selected from: (i) IL16, GROα and IL7; (ii) IL16, GROα and TNFβ; (iii) IL16, GROα, IL7, TNFβ and CTACK; and (iv) IL16, GROα, IL7, TNFβ, LIF and MIF.

4. The method according to claim 2, wherein the group of cytokines comprises IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9 and SCGFβ.

5. The method according to any one of claims 1-4, wherein the behavioral performance includes the scores of one or more screening tools or diagnostic tools for ASD.

6. The method according to claim 5, wherein the behavioral performance includes 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 or 13 scores selected from the group consisting of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT and SRS_MANN.

7. The method according to claim 6, wherein the behavioral performance includes a set of scores selected from: (i) ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; (ii) ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; (iv) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; and (v) ADOS_S, ADOS_C, ADOS_P, and CARS_total.

8. The method according to any one of claims 1 - 7, wherein the method comprises using a classifier to predict the response of a subject to bumetanide based on the feature information.

9. The method according to claim 8, wherein the classifier is selected from the group consisting of oblique random forest (ORF), partial least squares (PLS), sparse linear discriminant analysis (sLDA), neural network (NN), and support vector machine (SVM).

10. The method according to claim 8 or 9, wherein the classifier has been trained.

11. The method according to any one of claims 8 - 10, wherein: the feature information includes: (i) the baseline expression levels of IL16, GROα, IL7; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D; (iii) gender, age; and the classifier is a support vector machine; the feature information includes: (i) the baseline expression levels of IL16, GROα, TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, CARS_total; (iii) gender, age; and the classifier is partial least squares; the feature information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, and CTACK; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; (iii) gender and age; and the classifier is a neural network; the feature information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, LIF, MIF; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total; (iii) gender, age; and the classifier is sparse linear discriminant analysis; or the feature information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age; and the classifier is oblique random forest.

12. The method according to any one of claims 1-11, wherein the method comprises obtaining characteristic information of a subject by measuring the expression level of cytokines in a sample of the subject.

13. The method according to claim 12, wherein the sample is plasma.

14. A method for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, the method comprising: using a training data set comprising characteristic information of a plurality of ASD individuals and the determined response of the individuals to bumetanide to train a classifier, and selecting a subset of relevant characteristic information, wherein the characteristic information comprises: (i) the baseline expression levels of cytokines of IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9 and SCGFβ; (ii) the baseline scores of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT and SRS_MANN; and (iii) gender and age; predicting the response of the subject to bumetanide according to the subset of relevant characteristic information of the subject.

15. The method according to claim 14, wherein the classifier is selected from the group consisting of oblique random forest (ORF), partial least squares (PLS), sparse linear discriminant analysis (sLDA), neural network (NN) and support vector machine (SVM).

16. The method according to claim 14 or 15, wherein the expression level of the cytokines is obtained by measuring the expression level of the cytokines in a sample from the subject.

17. The method according to claim 16, wherein the sample is plasma.

18. A method for treating autism spectrum disorder (ASD) in a subject, the method comprising: using the method according to any one of claims 1-13 to predict the response of a subject with ASD to bumetanide; administering an effective amount of bumetanide to a subject identified as responsive to bumetanide.

19. A prediction device, comprising: an input module for receiving characteristic information of a patient with autism spectrum disorder (ASD), wherein the characteristic information comprises: (i) the baseline expression levels of a group of cytokines in the patient; (ii) the baseline behavioral manifestations of the patient; and (iii) the clinical information of the patient, including gender and age; a classification module comprising a classifier, wherein the classifier is capable of predicting the response of a subject to bumetanide according to the characteristic information using the classifier.

20. The prediction device according to claim 19, further comprising a training module configured to train a classifier using a training data set.

21. A computer-readable medium comprising computer-executable instructions recorded thereon for performing the following operations comprising: receiving characteristic information of a subject with autism spectrum disorder (ASD), wherein the characteristic information includes: (i) baseline expression levels of a set of cytokines of the subject; (ii) baseline behavioral manifestations of the subject; and (iii) clinical information of the subject, including gender and age; using a classifier algorithm based on the characteristic information to predict the response of the subject to bumetanide.

22. The computer-readable medium according to claim 21, wherein the operations further include training a classifier using a training data set.

23. The prediction device or computer-readable medium according to any one of claims 19-22, wherein the set of cytokines includes three or more cytokines selected from IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ.

24. The prediction device or computer-readable medium according to any one of claims 19-23, wherein the set of cytokines includes a group selected from (i) IL16, GROα, and IL7; (ii) IL16, GROα, and TNFβ; (iii) IL16, GROα, IL7, TNFβ, and CTACK; and (iv) IL16, GROα, IL7, TNFβ, LIF, and MIF.

25. The prediction device or computer-readable medium according to any one of claims 19-24, wherein the set of cytokines includes IL1β, IL6, IL8, IFNγ, TNFα, MCP1, Eotaxin, IL17, IL4, IL2Rα, MIG, MIP1β, IFNα2, SDF1α, IL16, LIF, TNFβ, MIF, RANTES, IL18, PDGFβ, IP10, IL13, MIP1α, GCSF, GROα, HGF, IL1α, SCF, TRAIL, MCSF, CTACK, IL7, IL9, and SCGFβ.

26. The prediction device or computer-readable medium according to any one of claims 19-25, wherein the behavioral performance includes 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, or 13 scores selected from the group consisting of CAR_total, CAR_S, CAR_N, CAR_D, ADOS_S, ADOS_C, ADOS_P, ADOS_I, SRS_total, SRS_AWA, SRS_COG, SRS_COM, SRS_MOT, and SRS_MANN.

27. The prediction device or computer-readable medium according to any one of claims 19-26, wherein the behavioral performance includes a set of scores selected from the following: (i) ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; (ii) ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; (iv) ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; and (v) ADOS_S, ADOS_C, ADOS_P, and CARS_total.

28. The prediction device or computer-readable medium according to any one of claims 19-27, wherein the classifier is selected from the group consisting of oblique random forest (ORF), partial least squares (PLS), sparse linear discriminant analysis (sLDA), neural network (NN), and support vector machine (SVM).

29. The prediction device or computer-readable medium according to any one of claims 19-28, wherein the classifier has been trained.

30. The prediction device or computer-readable medium according to any one of claims 19-29, wherein: the feature information includes: (i) the baseline expression levels of IL16, GROα, and IL7; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, and CARS_D; (iii) gender and age; and the classifier is a support vector machine; the feature information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age; and the classifier is partial least squares; the feature information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, and CTACK; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, CARS_total, and SRS_MOT; (iii) gender and age; and the classifier is a neural network; The characteristic information includes: (i) the baseline expression levels of IL16, GROα, IL7, TNFβ, LIF, and MIF; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, SRS_COM, CARS_D, and CARS_total; (iii) gender and age; the classifier is sparse linear discriminant analysis; or The characteristic information includes: (i) the baseline expression levels of IL16, GROα, and TNFβ; (ii) the scores of ADOS_S, ADOS_C, ADOS_P, and CARS_total; (iii) gender and age; the classifier is oblique random forest.

31. A kit for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, the kit comprising reagents for measuring the expression levels of a set of cytokines defined in any one of claims 1-13 and a specification describing the prediction method according to any one of claims 1-13.

32. A kit for treating autism spectrum disorder (ASD) in a subject, the kit comprises: agents for measuring the expression levels of a set of cytokines defined in any one of claims 1-13; bumetanide; and a specification describing the prediction method according to any one of claims 1-13, and administering an effective amount of bumetanide to a subject identified as responsive to bumetanide.

33. The characteristic information defined in any one of claims 1-13, for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide.

34. Use of the characteristic information defined in any one of claims 1-13 to prepare a set of characteristics for predicting the response of a patient with autism spectrum disorder (ASD) to bumetanide.

35. Use of an agent for measuring the expression levels of the cytokine group defined in any one of claims 1-13 in the preparation of an agent or a kit for predicting the response of a subject with autism spectrum disorder (ASD) to bumetanide, wherein the prediction is carried out by the prediction method according to any one of claims 1-13.

36. Establishing a prediction model using the characteristic information defined in any one of claims 1 to 13 for predicting the response of a patient with autism spectrum disorder (ASD) to bumetanide.

Images