Utilizing principal component analysis to identify clinically relevant changes in human leukocyte antigen testing patterns

Abstract

A method for analyzing human leukocyte antigen (HLA) test results to identify clinically relevant changes in test patterns or evaluate HLA testing includes receiving, as input, HLA test results. The method further includes using principal component analysis (PCA) to analyze the HLA test results to detect a clinically relevant change for a subject or evaluate an aspect of HLA testing. The method further includes outputting results of the analysis.

Description

[0001] DESCRIPTION

[0002] METHODS, SYSTEMS, AND COMPUTER READABLE MEDIA FOR UTILIZING PRINCIPAL COMPONENT ANALYSIS (PC A) TO IDENTIFY CLINICALLY RELEVANT CHANGES IN HUMAN LEUKOCYTE ANTIGEN (HLA) SINGLE ANTIGEN BEAD SOLID PHASE TESTING PATTERNS

[0003] CROSS REFERENCE TO RELATED APPLICATION

[0004] The presently disclosed subject matter claims the benefit of U.S. Provisional Patent Application Serial No. 63 / 525,149, filed uly 5, 2023, the disclosure of which incorporated herein by reference in its entirety.

[0005] TECHNICAL FIELD

[0006] The subject matter described herein relates to analyzing HLA test results. More particularly, the subject matter described herein relates to using principal component analysis (PCA) to identify clinically relevant changes in patient HLA single antigen bead solid phase testing patterns.

[0007] BACKGROUND

[0008] Human leukocyte antigen (HLA) antibody testing is essential for successful solid-organ allocation, patient monitoring post-transplant, and risk assessment for both solid-organ and hematopoietic transplant patients. LUMINEX® solid-phase testing is the most common method for identifying HLA antibody specificities, making it one of the most complex immunoassays as each panel contains over 90 specificities for both HLA class I and HLA class II with most of the analysis being performed manually in the vendor-provided software. Principal component analysis (PCA), used in machine learning, is a feature extraction method often utilized to assess data with many variables.

[0009] SUMMARY

[0010] This Summary lists several embodiments of the presently disclosed subject matter, and in many cases lists variations and permutations of these embodiments. This Summary is merely exemplary of the numerous and varied embodiments. Mention of one or more representative features of a given embodiment is likewise exemplary. Such an embodiment can typically exist with or without the feature(s) mentioned; likewise, those features can be applied to other embodiments of the presently disclosed subject matter, whether listed in this summary or not. To avoid excessive repetition, this Summary does not list or suggest all possible combinations of such features. In some embodiments, the presently disclosed subject matter relates to methods for analyzing human leukocyte antigen (HLA) test results to identify clinically relevant changes in test patterns and / or to evaluate HLA testing. In some embodiments, the presently disclosed methods comprise receiving, as input, HLA test results. In some embodiments, the presently disclosed methods further comprise using principal component analysis (PC A) to analyze the HLA test results to detect a clinically relevant change for a subject and / or to evaluate an aspect of HLA testing. In some embodiments, the methods further comprise outputting results of the analysis.

[0011] In some embodiments, the presently disclosed subject matter relates to methods for evaluating performance of different sets of otherwise equivalent HLA testing reagents. In some embodiments, the performance of a second set of HLA testing reagents is compared to the performance of a first set of HLA testing reagents with respect to the ability to provide HLA testing data. In some embodiments, the methods comprise, consist essentially of, or consist of receiving, as input, HLA test results from the second set of HLA testing reagents; using principal component analysis (PCA) to analyze the HLA test results from the second set of HLA testing reagents; and comparing the HLA test results from the second set of HLA testing reagents to equivalent HLA test results from the first set of HLA testing reagents, wherein a difference between the HLA test results from the second set of HLA testing reagents and the HLA test results from the first set of HLA testing reagents is indicative of a difference in performance of the second set of HLA testing reagents as compared to the first set of HLA testing reagents. In some embodiments, the presently disclosed methods further comprise outputting results of the analysis. In some embodiments, the outputting comprises displaying the results of the analysis on a viewing device (e.g., a monitor, a cellular telephone, etc.). In some embodiments, the comparing is indicative of the second set of HLA testing reagents performing less effectively than the first set of HLA testing reagents with respect to identifying presence of donor specific antibodies (DS As) in a biological sample isolated from a subject. In some embodiments, first and second sets of HLA reagents comprise HLA class I beads, HLA class II beads, or a combination thereof.

[0012] In some embodiments, the presently disclosed subject matter also relates to methods for assessing HLA antibody testing. In some embodiments, the methods comprise, consist essentially of, or consist of receiving, as input, a first set of HLA test results; using principal component analysis (PCA) to analyze the first set of HLA test results; and comparing the first set of HLA test results to a second set of HLA testing results, wherein the first and second set of HLA testing results are derived from testing of a biological sample with a first and second set of HLA testing reagents, respectively, under equivalent conditions, wherein a difference between the first and second sets of HLA testing is indicative of a difference in performance of the first and second sets of HLA testing reagents. In some embodiments, the presently disclosed methods further comprise outputting results of the analysis, wherein the outputting optionally comprises displaying the results of the analysis on a viewing device (e.g., a monitor, a cellular telephone, etc.). In some embodiments, the comparing is indicative of the second set of HLA testing reagents performing less effectively than the first set of HLA testing reagents with respect to identifying presence of donor specific antibodies (DSAs) in the biological sample. In some embodiments, the first and second sets of HLA reagents comprise HLA class I beads, HLA class II beads, or a combination thereof.

[0013] In some embodiments of the presently disclosed methods, the HLA test results are generated by HLA testing of a biological sample isolated from a subject, and further wherein the subject is a pre-transplant subject or a post-transplant subject. In some embodiments, the HLA test results comprise HLA test data comprising HLA test data related to the presence of donor specific antibodies (DSAs) in the subject, optionally wherein the presence of DSAs was identified in samples for the subject before and / or after a solid organ or a hematopoietic transplant into the subject.

[0014] In some embodiments, the presently disclosed subject matter relates to non- transitory computer readable media having stored thereon executable instructions that when executed by a processor of a computer control the computer to perform the method as claimed in any of the preceding claims. Thus, in some embodiments the subject matter described herein can be implemented in software in combination with hardware and / or firmware. For example, the subject matter described herein can be implemented in software executed by a processor. In some embodiments, the subject matter described herein can be implemented using a non-transitory computer readable medium having stored thereon computer executable instructions that when executed by the processor of a computer control the computer to perform steps. Exemplary computer readable media suitable for implementing the subject matter described herein include non-transitory computer-readable media, such as disk memory devices, chip memory devices, programmable logic devices, and application-specific integrated circuits. In addition, in some embodiments a computer readable medium that implements the subject matter described herein can be located on a single device or on a computing platform or can be distributed across multiple devices or computing platforms.

[0015] Accordingly, it is an object of the presently disclosed subject matter to provide methods for analyzing human leukocyte antigen (HLA) test results to identify clinically relevant changes in test patterns and / or to evaluate HLA testing.

[0016] These and other objects are achieved in whole or in part by the presently disclosed subject matter. Further, objects of the presently disclosed subject matter having been stated above, other objects and advantages of the presently disclosed subject matter will become apparent to those skilled in the art after a study of the following detailed description, Figures, and EXAMPLES.

[0017] BRIEF DESCRIPTION OF THE FIGURES

[0018] Exemplary implementations of the subject matter described herein will now be explained with reference to the accompanying Figures, of which:

[0019] Figures 1A-1C. Principal component analysis of HLA class I results demonstrated that samples with donor specific antibodies (DSA or DSAs) had higher Euclidean distances. (Figure 1 A) For patients with class I DSAs, the average Euclidean distance observed was higher in DSA samples compared to pre-DSA samples. (Figure IB) Representative PCA dot plots for PCI and PC2 are displayed for DSA, highly sensitized pre-transplant, and low reactive post-transplant sample cases. (Figure 1C) The fold difference observed between the average Euclidean distance of each sample type and the overall average Euclidean distance between patient samples was compared. ***p-value < 0.005, ****p-value <0.0001.

[0020] Figures 2A-2C. Principal component analysis of HLA class II results demonstrated that samples with class II DSAs had higher Euclidean distances. (Figure 2A) For patients with class II DSAs, the average Euclidean distance observed was higher in DSA samples compared to pre-DSA samples. (Figure 2B) Representative PCA dot plots for PCI and PC2 are displayed for DSA, highly sensitized pre-transplant, and low reactive post-transplant sample cases. (Figure 2C) The fold difference observed between the average Euclidean distance of each sample type and the overall average Euclidean distance between patient samples was compared. ***p-value < 0.005, ****p- value <0.0001. Figures 3A and 3B. Fold differences in Euclidean distance values can differentiate samples with HLA antibody pattern deviations from patient history. (Figure 3 A) Receiver operator characteristic analysis performed for HLA class I single antigen bead results, demonstrating an area under the curve of 0.9535. (Figure 3B) Receiver operator characteristic analysis performed for HLA class I single antigen bead results, demonstrating an area under the curve of 0.9762.

[0021] Figures 4A-4D. Principal component analysis of HLA class I and class II beads using a curated data set across multiple bead lots, identified single antigen beads which displayed potential over reactivity. (Figure 4A) Principal component analysis of class I single antigen beads, highlighting beads with the greatest Euclidean distance from all other beads. (Figure 4B) Principal component analysis of class II single antigen beads, highlighting beads with the greatest Euclidean distance from all other beads. (Figure 4C) Heat map demonstrating the Euclidean distance calculated post PCA for all class I beads. (Figure 4D) Heat map demonstrating the Euclidean distance calculated post PCA for all class II beads.

[0022] Figure 5 is a block diagram illustrating an exemplary system for implementing the subject matter described herein.

[0023] Figure 6. Analysis of a cohort of patient samples with HLA Class II bead over reactivity. Principal component analysis of a subset of patient samples with suspected DPAl*02:01-DPBl*01 :01 and DPAl*02:02-DPBl*05:01 over reactivity were analyzed. PCA derived analysis revealed that the method could distinguish the suspected overreactive beads from all of the other beads from the class II single antigen bead panel.

[0024] DETAILED DESCRIPTION

[0025] HLA antibody testing first became recognized as clinically relevant in transplant medicine in the 1960s (Patel & Terasaki, 1969) and since then studies for all transplanted solid-organs have found that monitoring for anti-HLA antibodies is important for patient clinical care and prognosis (Wiebe et al., 2012; Verleden et al., 2017). Clinical testing most widely used for detecting HLA antibodies are solid phase multiplex-based assays. Serum test results using these assays influence virtual crossmatch estimations, identification of unacceptable antigens, and characterization of donor-specific antibodies (Amico et al., 2009; Yanagida et al., 2011).

[0026] Machine learning algorithms often use principal component analysis (PCA) and resultant matrices while processing input data sets. This is done as data sets analyzed by machine learning algorithms are often extremely large and cumbersome. PCA, as a data reduction technique, assists with extracting the most influential variables of a given data set. PCA is used in a wide variety of applications including image analysis software, such as facial recognition, finance evaluation, data exploration, and cyber security risk assessment (Elmore & Richman, 2001; Zhi & Liu, 2019; Garate-Escamila et al., 2020).

[0027] LUMINEX®-based solid-phase HLA antibody testing is a complex test performed in the clinical histocompatibility laboratory. Both HLA class I and HLA class II antibody assays can detect over 96 antibody specificities each with more bead specificities that are becoming available and employed in patient testing. When comparing patient historic antibody test results to current results, many medical laboratory scientists utilize the recommended company-provided software packages to visually examine samples and determine clinical significance based on their transplant center and laboratory standards of practice. The presently disclosed study evaluated the application of a dimensionality reduction technique, PCA, to approach data analysis as a proof of concept. The goal was to determine if PCA-assisted data analysis could help in identifying clinically relevant changes in patient samples. Additionally, the analysis described herein explored the use of PCA in identifying potentially over-reactive single antigen beads from a large transplant center data cohort.

[0028] The subject matter described herein can be implemented in software in combination with hardware and / or firmware. For example, the subject matter described herein can be implemented in software executed by a processor.

[0029] Figure 5 is a block diagram illustrating an exemplary system for implementing the subject matter described herein. In Figure 5, the system includes computing platform 500 having at least one processor 501 and memory 502. The system further includes HLA testing data analyzer 504 for performing the steps described herein for analyzing HLA testing data 503 to identify clinically relevant changes for a subject or to evaluate aspects of the HLA testing, such as reactivity differences between bead lots. HLA testing data analyzer 504 may be implemented using computer executable instructions stored in memory 502 and executed by processor 501.

[0030] The presently disclosed subject matter can be employed for various uses related to HLA testing in addition to those disclosed herein above. For example, in some embodiments the presently disclosed subject matter relates to methods for evaluating performance of different sets of otherwise equivalent HLA testing reagents. In such embodiments, the presently disclosed methods can be used for performing quality assurance / quality control assessments of various reagents employed for HLA testing.

[0031] By way of example and not limitation, in some embodiments the presently disclosed subject matter can be used to compare the performance of a one set of HLA testing reagents to the performance of another set of HLA testing reagents with respect to the ability to provide HLA testing data. In such embodiments, the first and second HLA testing reagents are designed to perform similarly or identically, but the ability of the first and second testing reagents to do so can vary. To determine whether the first and second HLA testing reagents do indeed perform similarly or identically, the presently disclosed methods can be employed to compare the performance of the two sets of HLA testing reagents.

[0032] Thus, in some embodiments, the presently disclosed methods can comprise, consist essentially of, or consist of receiving, as input, HLA test results from one set of HLA testing reagents, using principal component analysis (PCA) to analyze the HLA test results from this set of HLA testing reagents, and comparing the HLA test results from this set of HLA testing reagents to HLA test results from another set of ostensibly equivalent HLA testing reagents, wherein a difference between the HLA test results from the two sets of HLA testing reagents is indicative of a difference in performance between the two sets of HLA testing reagents. In some embodiments, the two sets of HLA testing reagents are different lots of the same form of HLA testing reagents (e.g., two different lots of HLA testing beads).

[0033] Furthermore, the presently disclosed subject matter also relates to methods for assessing HLA antibody testing, which in some embodiments can comprise, consist essentially of, or consist of receiving, as input, a first set of HLA test results; using principal component analysis (PCA) to analyze the first set of HLA test results; and comparing the first set of HLA test results to a second set of HLA testing results, wherein the first and second set of HLA testing results are derived from testing of a biological sample with a first and second set of HLA testing reagents, respectively, under equivalent conditions, wherein a difference between the first and second sets of HLA testing is indicative of a difference in performance of the first and second sets of HLA testing reagents. Thus, in some embodiments the presently disclosed methods can be employed for comparing multiple (e.g., two) sets of HLA testing reagents to determine if they perform acceptably for HLA testing. By way of example and not limitation, the presently disclosed methods can be employed for determining if one set of HLA testing reagents perform less effectively than another set of HLA testing reagents with respect to identifying presence of donor specific antibodies (DSAs) in the biological sample.

[0034] In some embodiments of the presently disclosed methods, the HLA reagents that are compared comprise HLA class I beads, HLA class II beads, or a combination thereof, such as those disclosed herein.

[0035] In some embodiments of the presently disclosed methods, the HLA test results are generated by HLA testing of a biological sample isolated from a subject, and further wherein the subject is a pre-transplant subject or a post-transplant subject. In some embodiments, the HLA test results comprise HLA test data comprising HLA test data related to the presence of donor specific antibodies (DSAs) in the subject, optionally wherein the presence of DSAs was identified in samples for the subject before and / or after a solid organ or a hematopoietic transplant into the subject. In some embodiments, the presently disclosed methods can be employed for determining whether there is a change in a DSA profile (i.e., the set of DSAs present in an individual at one time as compared to another). The DSA profile can include data with respect to DSAs, including but not limited to historical de novo DSAs (dnDSAs), DSAs that were present before a subject transplant, or any combination thereof.

[0036] In some embodiments, the presently disclosed methods further comprise, consist essentially of, or consist outputting the results of the analysis. In some embodiments, the outputting comprises displaying the results of the analysis on a viewing device (e.g., a monitor, a cellular telephone, etc.). In some embodiments, the comparing is indicative of the second set of HLA testing reagents performing less effectively than the first set of HLA testing reagents with respect to identifying presence of donor specific antibodies (DSAs) in a biological sample isolated from a subject. In some embodiments, first and second sets of HLA reagents comprise HLA class I beads, HLA class II beads, or a combination thereof.

[0037] EXAMPLES

[0038] Materials and Methods for the EXAMPLES

[0039] Samples. Patient samples were selected from solid organ pre-transplant, solid organ post-transplant, and hematopoietic stem cell transplant patients. All patient samples were collected and tested between January 2018 and March 2022 at McLendon Laboratories, UNC Hospital, Chapel Hill, North Carolina, United States of America. Patient data and sample results were collected per the UNC internal review board (IRB)-approved study parameters. The samples included in the proof-of-principle runs using PCA with our established thresholds were deidentified after all data had been compiled and analyzed. This was done because the clinical results needed to be extracted and paired with the PCA analysis for predicting clinically impactful changes in HLA antibodies. All other samples were deidentified after data collection and before PCA analysis.

[0040] Solid Phase Testing. Solid phase testing for HLA antibodies was performed using HLA single antigen bead class I and class II assays (One Lambda; Thermo Fisher Scientific Inc., Waltham, Massachusetts, United States of America) on the FLEXMAP 3D® platform (LUMINEX® Corporation, Austin, Texas, United States of America). Sample testing was performed using patient serum and tested as recommended by the assay manufacturer (Bray et al., 2011; Gandhi et al., 2013).

[0041] Bioinformatics. Machine learning algorithms often utilize matrices of covariance to identify differences between occurrences, images, or patterns that have multiple variables, in our study we utilized a similar approach to investigate MFI values from HLA solid phase single antigen bead results. Patient results for solid phase single antigen bead testing and sample identifiers were obtained using the FUSION™ software (v4.2, One Lambda; Thermo Fisher Scientific Inc., Waltham, Massachusetts, United States of America). MFI values were obtained by exporting all test samples for the given time to .csv files. These .csv files were then concatenated utilizing R. The pre-transplant control subjects, post-transplant control subjects, and patients with donor-specific antibodies (DSA) were all de-identified and assigned a study number. All samples were assigned a sample study number per IRB protocol. Patient sample results were compiled, de-identified, and quality control samples removed using R (version 4.1.1). Analysis and grading of sample results based on the Euclidean distance was performed blinded from the review and clinical resulting process.

[0042] Statistical Analysis. Principal component analysis was performed using GraphPad Prism (version 9.4.1; GraphPad Software, Boston, Massachusetts, United States of America) and R (version 4.1.1). Euclidean distance analysis based on principal component values was performed using the equation, Euclidean distance (ED) = / (x1- x2)2+ (r1- r2)2, as has previously been established and described for the analysis of principal component-derived values (Sari et al., 2018; Kuncoro & Sugiharti, 2021). The distance ratio of samples was calculated as the Euclidean distance of the sample of interest divided by the average Euclidean distance of the other patient samples. Receiver operating characteristic (ROC) analysis was performed using GraphPad Prism (version 9.4.1; GraphPad Software, Boston, Massachusetts, United States of America). The paired t-test or Mann-Whitney U test was performed when appropriate. For all statistical tests, calculated P-values < 0.05 were considered significant.

[0043] Introduction to the EXAMPLES

[0044] Machine learning algorithms utilize matrices of covariance to identify differences between occurrences, images, or patterns that have multiple variables. This is often referred to as the principal component analysis (PCA) method which utilizes the variance across multiple variables of multiple training samples to show similarity or dissimilarity between compared patterns from future samples (Kherif & Latypova, 2020). In the PCA method, the Eigenvectors are calculated from the covariance matrix and then the unique features are displayed on a lower dimensional plane, typically with each plane, referred to as a principal component, accounting for a percentage of the variance observed across the samples compared (Novembre et al., 2008; Kherif & Latypova, 2020). One way to determine if the pattern is similar or dissimilar is to measure the distance on the resultant planes used between the reference samples and the sample in question. The distance is referred to as the Euclidean distance if this is calculated on a two-dimensional plane (Elmore & Richman, 2001). In some embodiments, the presently disclosed subj ect matter was designed to evaluate the utility of PCA in identifying clinically significant differences in solid-phase HLA antibody testing results using this method.

[0045] EXAMPLE 1

[0046] Principal Component Analysis of HLA Class I Results Demonstrated that Samples with PSA Had Higher Euclidean Distances

[0047] To initially evaluate if this method could be appropriately applied to HLA antibody results, 25 patients that were positive for HLA donor-specific antibodies (DS A) between January 2019 and December 2021 were selected. PCA analysis was performed on each patient’s historic pre-DSA samples and the sample in which DSA was identified. Samples with DSA present a difference in pattern compared to the patient’s previous historic samples and are clinically significant. The average Euclidean distance was calculated post-PCA analysis between each of the historic samples. This was then compared to the average Euclidean distance exhibited between the DSA sample and each of the historic samples. This was performed for HLA class I (Figure 1 A) DSAs. We found that the Euclidean distance was significantly higher for the DSA samples when they were compared to the historic pre-DSA sample Euclidean distances for our patient cohort for HLA class I (paired t-test, p < 0.0005). To further evaluate if this method could distinguish this known reactivity difference compared to samples that displayed a similar antibody pattern, two control cohorts of patient samples were also compared. Sensitized pre-transplant patients, comprising 159 patient samples from 13 different patients, were utilized as one control group. The second control group was composed of post-transplant patient samples that lacked reactivity, totaling 163 patient samples among 20 different patients.

[0048] PCA analysis was performed to compare samples for each patient in both cohorts. Then the Euclidean distance was calculated for HLA class I samples. Examples of how the PCA analysis visually appears for select patients for the DSA cohort and the two control cohorts are displayed for HLA class I (Figure IB). To determine if there was a significant difference between sample Euclidean distances post PCA analysis among the cohort with DSA and the two control cohorts, the fold difference between each sample and the average Euclidean distance observed for each patient was calculated for HLA class I (Figure 1C). Our cohort of patients with DSAs was further sub-grouped by whether the DSA identified was historical de novo DSA (dnDSA) or was DSA that was present before the patient’s transplant.

[0049] EXAMPLE 2

[0050] Principal Component Analysis of HLA Class II Results Demonstrated that Samples with DSA Had Higher Euclidean Distances

[0051] A similar approach as described in EXAMPLE 1 for HLA class I samples was followed for HLA class II samples. It was determined that the Euclidean distances was significantly higher for the DSA samples when they were compared to the historic pre- DSA sample Euclidean distances for our patient cohort for HLA class II (Figure 2A, paired t-test, p < 0.0005). Examples of how the PCA analysis visually appears for several patients with DSA and the two control cohorts are displayed for HLA class II (Figure 2B). For HLA class II the average Euclidean distance observed for each patient was calculated as described for the HLA class I patient samples (see Figure 1C).

[0052] It was observed that the Euclidean distance differences between samples post- PCA, expressed as fold change, was similar for samples that did not have drastically different changes in HLA class I (Mann-Whitney U, p = 0.0849) and HLA class II (Mann-Whitney U, p = 0.920) antibody profiles (Figures 1C and 2C, respectively). The fold change was calculated as the distance of the current sample over the average distance from all the other samples from the same patient, thus providing a ratio of how much more distance the sample being assessed exhibited in comparison to all other samples from the same patient that were considered. However, samples that had dnDSA that were identified in post-transplant testing, were more distant upon comparison using PCA, as expressed as a fold difference, in single antigen bead testing for both HLA class I (Mann-Whitney U, p < 0.0001) and HLA class II (Mann-Whitney U, p < 0.0001). The same was found for both HLA class I (Mann-Whitney U, p < 0.0001) and HLA class II (Mann-Whitney U, p < 0.0001) samples where the DSA identified was historic and not dnDSA.

[0053] EXAMPLE 3

[0054] Fold Differences in Euclidean Distance Values Can Differentiate Samples with HLA Antibody Pattern Deviations from Patient History

[0055] Based on the findings described herein above, it seemed likely that a threshold could be established for the fold difference in Euclidean distance that could distinguish a clinically significant change in patient antibody results from PCA analysis from those that exhibited similar antibody patterns with no clinically significant change. To establish this cut-off value for both HLA class I and HLA class II antibody testing, ROC analysis was performed using the pre-transplant and post-transplant patient cohort Euclidean distance ratios as negative controls and the Euclidean distance ratios from the patient samples with dnDSAs and DSAs as our positive control samples.

[0056] The results are presented in Figures 3 A and 3B . The area under the curve (AUC) was calculated for HLA class I (see Figure 3A; AUC = 0.954) and HLA class II (see Figure 3B; AUC = 0.976) samples. ROC analysis found that for both HLA class I and HLA class II single antigen tests, a Euclidean distance ratio of 1.4 was optimal to distinguish the samples with DSA from the samples which have similar patterns of expression but were not clinically impactful.

[0057] To evaluate if this threshold value for HLA class I and HL A class II could autonomously identify samples with clinically significant changes in DSA antibody testing, pilot samples were tested using blinded PCA analysis on newly received samples for DSA antibody testing by the HLA class I and HLA class II single antigen bead assay. One HLA class I and one HLA class II batch were selected for the pilot study, totaling 24 HLA class I and 25 HLA class II samples, respectively. After antibody testing was performed, the normalized MFI values were exported from the FUSION™ software (v4.2, One Lambda; Thermo Fisher Scientific Inc., Waltham, Massachusetts, United States of America) before analysis (or manual review) and analyzed using the established PCA method. Samples lacking reactivity, all demonstrating bead MFI values below the assay cut-off (MFI < 1000) for positivity, were discarded from the pilot study, resulting in a total of 17 discarded samples. In total, 16 HLA class I samples and 16 HLA class II samples were included from the single antigen bead runs.

[0058] Utilizing the derived principal component scores for the samples, the Euclidean distance ratio was calculated. If the Euclidean distance ratio was equal to or above 1.4, it was flagged as worthy of clinical investigation and likely abnormal from historic patient sample results. If the value was less than 1.4, the sample was not flagged and considered normal according to previous sample reactivity. These PCA results were then blindly compared to the sample's clinically reported results which receive two manual reviews by laboratory staff and a final review by the laboratory director. The utility of PCA in screening samples for atypical HLA antibody patterns is displayed in Table 1.

[0059] Table 1

[0060] Performance of PC A-Derived Analysis for SAB Class I and II

[0061] * PPV: positive predictive value. NPV: negative predictive value PCA analysis on these samples demonstrated excellent sensitivity (100%, 95% CI = 73.54% - 100%) in identifying samples that were truly abnormal and dissimilar to patient historic antibody testing (negative predictive value, 100%). Of note, PCA analysis identified all of the samples which were positive for DSA but were not de novo (4 / 4), one sample which had de novo DSA (1 / 1), and all of the samples which had irregular reactivity (7 / 7) compared to previous patient HLA antibody history as noted by the clinical HLA laboratory staff members requiring repeat testing, additional treatment or investigation, including one sample that exhibited complement dependent ‘prozone’ activity. The specificity for this methodology was 78.26% (95% CI = 56.30% - 92.54%) and was likely due to multiple samples for both HLA class I and HLA class II that were flagged as abnormal although they did not have DSAs and were not considered clinically different from the patients previously tested historic samples. These samples were investigated, and their details are provided in Table 2. While all of these samples lacked new reactivity, they did differ from previous samples because of an increase in MFI values in one or more of the HLA specificities. The baseline MFI values for each of the specificities from these samples exhibited an increase of 2-fold or more compared to the average baseline MFI values from previous historic tests.

[0062] Table 2

[0063] False Positive Sample Reactivity

[0064] EXAMPLE 4

[0065] Utility of PCA in Identifying Single Antigen Beads that Might Exhibit OverReactivity for the HLA Class I and HLA Class II Solid Phase Assays

[0066] As PCA is a method of identifying variance across multiple variables, the utility of PCA in identifying single antigen beads that might exhibit over-reactivity for the HLA class I and HLA class II solid phase assays was investigated. To investigate if PCA would appropriately identify potentially false positive beads, a cohort of patient samples was selected that had known or suspected over-reactivity for the DPAl*02:01~DPBl*01:01 and the DPAl*02:02~DPBl*05:01 beads, comprising 121 samples. These samples had baseline MFI values equal to or less than 1,200. PCA was performed using these samples with each sample utilized as a loading variable. If all beads had similar variation across samples then it would be expected that the beads would have similar PC scores and localize to a similar region of the PCA plot.

[0067] The analysis revealed that while nearly all the beads exhibited similar PC scores for the two most significant eigenvectors, the DPAl*02:01~DPBl*01 :01 and the DPAl*02:02~DPBl*05:01 beads exhibited a Euclidean distance that was 6.2 and 4.7 fold, respectively, more distant on average from all the other HLA class II beads (see Figure 6). Further, as expected since these samples were selected for their distinct pattern of reactivity, the DPA1 *02:0 l-DPB 1*01 :01 and the DPAl*02:02~DPBl*05:01 beads aligned along similar loading vectors and localize to a similar position on the PCA plot such that, the PC values for these two beads are more similar to each other than to the other single antigen bead PC values for HLA class II.

[0068] EXAMPLE 5

[0069] PCA of HLA Class I and Class II Beads Using a Curated Data Set Across Multiple Bead Lots Identified Single Antigen Beads That Displayed Potential Over Reactivity

[0070] After verifying that this method was appropriate in identifying beads with suspected over-reactivity in the pilot data set described herein above, the utility of this approach across all samples tested at the UNC HLA laboratory from 2018 through March 2022 for HLA class I and HLA class II was investigated. The results for HLA class I and HLA class II are presented in Figures 4A and 4B, respectively.

[0071] Samples were included in the analysis if they had baseline MFI values less than 1,000. In total, there were 3,330 samples included in the analysis originally tested using the HLA class I single antigen bead assay and 3,288 samples tested using the HLA class II single antigen bead assay. Testing identified several outlying beads for HLA class I and HLA class II, representing possible overreactive beads observed commonly in our population of patients tested since 2018. For each bead, the average Euclidean distance ratio was calculated, and each bead was ranked based on its distance from the average distance of all beads on the PC plot using PCI and PC2 for both HLA class I and HLA class II (see Tables 3 and 4). The HLA class I single antigen bead assay contains some beads that might be overreactive including B*15: 12, C*12:03, and the A29 beads (Table 3). Several potentially overreactive beads in the HLA class II single antigen bead assay included the DRB3*02:02, DPAl*02:01-DPBl*01:01, DPAl*02.02- DPBl*05:01, DQAl*05:05-DQBl*03:01, DRBl*04:01, and DRBl*04:03 beads among others (Table 4).

[0072] Table 3

[0073] Class I Bead PCA Euclidean Distance Comparison

[0074] Table 4 Class II Bead PCA Euclidean Distance Comparison

[0075] Discussion of the EXAMPLES

[0076] Multiplex solid-phase assays for the detection of HLA antibodies have transformed the field of HLA diagnostics and transplant medicine. Some of the key advantages of this method are the increased sensitivity and specificity for detecting HLA antibodies compared to cytotoxicity and other cell-based assays. As a result of these assays, there has been a sweeping change in tools that highly influence organ allocation, such as the ‘virtual crossmatch’ and ‘calculated panel reactive antibodies’ which have become essential components for center-specific immunological risk stratification for both solid and stem cell transplant, as well as kidney-paired donation programs (Wehmeier et al., 2020; Mamode et al., 2022; Yu et al., 2022). It is the vital task of the transplant immunology laboratory to identify these HLA antibodies and convey the risk associated with the laboratory findings to the clinical care teams. As the task of identifying HLA antibodies becomes ever more complex, the field would greatly benefit from additional bioinformatic pipelines and software tools to assist laboratory staff in the post-analytical phase of testing (Weimer & Newhall, 2019).

[0077] As many solid organs and hematopoietic stem cell transplant patients exhibit a rather complex pattern of HLA antibodies due to various sensitizing events throughout their health care and life experiences, a computer-assisted analysis of these patterns would be ideal. Utilized for variance and pattern analysis by facial recognition software, fiscal evaluation, cyber security risk assessment, and machine learning algorithms, principal component analysis (PCA) presents a possible method for providing analysis of HLA antibody patterns, particularly those with more complex HLA antibody profiles. PCA is a dimensionality reduction technique, which uses feature extraction rather than feature exclusion, making it useful to compare variance among samples across many variables. Our study found that PCA analysis was able to enhance automated HLA antibody analysis for individual patients and assessment of bead over reactivity across a database of patient results.

[0078] Recent publications described the use of PCA to assess HLA single antigen bead data collected across three different histocompatibility centers (Vittoraki et al., 2020; Vittoraki et al., 2021). In the adult population that was investigated using principal component analysis, it was found that this method could identify “patterns in human immune responses with striking similarities with the previously described CREGs” (Vittoraki et al., 2020). It was demonstrated that anti-DP antigenic responses did not seem to be correlated to anti -DR or -DQ responses, while the latter two did tend to show a correlation based on PCA analysis (Vittoraki et al., 2020). In a study published a year later, the same group described how through the use of PCA in their particular machine learning algorithm, it was possible to identify several previously identified crossreactive groupings and additional new crossreactive patterns of alloreactivity (Vittoraki et al., 2021). A helpful feature of these studies gave a detailed analysis of all bead antigenic specificities, represented as a dendrogram, calculating a measure of allelic antigenic distances for these bead-array-defined cross-reactive groups (Vittoraki et al., 2021).

[0079] The presently disclosed study utilized PCA on a more laboratory application- focused scale to preemptively analyze 24 patient HLA class I and 25 patient HLA class II HLA DSA antibody testing samples and identify samples that had deviant patterns from each patient’s historic antibody profile based on their previous testing results. Using the Euclidean distance ratio derived from principal component scores and the defined threshold which would predict a deviant antibody pattern, the PCA-derived results was compared to the laboratory staff review, testing outcomes, and finalized patient results for the HLA DSA antibody tests performed. PCA-derived analysis was able to identify all samples which had post-transplant DSA (4 / 4) and all other samples which had noted irregular patterns of reactivity (8 / 8), including one sample that exhibited a complement-mediated prozone effect. PCA had a sensitivity of 100.00% (95% CI, 73.54%-100.00%) with a negative predictive value of 100.00% across both patient cohorts.

[0080] The evaluated specificity was slightly lower at 78.26% (95% CI, 56.30%- 92.54%) with a positive predictive value of 70.56% (95% CI, 52.50%-83.90%). This lower positive predictive value was due to five patient samples that PCA-derived analysis identified as deviant from historic samples, but which laboratory staff resulted with no abnormal review remarks, comments, repeat testing, or DSA and which would have been considered consistent with history by staff with no clinical impact. Each of these patient samples was investigated and compared to their historic samples. The results of these comparisons are displayed in Table 2. While there was no large visual difference when looking at the patient antibody histograms and all antibody specificities were in the same order when organized from most reactive to least reactive, there was a difference in the MFI values calculated in these five samples. In each of the five samples, the HLA specificities that historically had been above 1,000 MFIs, all exhibited at least a doubling of their MFI values, with one HLA class II sample being 243% more reactive than the average MFI values measured from the patient’s historic samples. While this increase in sample reactivity might not be clinically impactful in these five cases, cases like these must be identified and investigated as a significant increase in MFI values, as exhibited by these samples, could be clinically impactful in certain instances.

[0081] These results provided evidence that PCA-assisted analysis, given the excellent negative predictive value and adequate positive predictive value, could be utilized in assisting laboratory analysis of patient samples. Depending on the laboratory workflow, this analysis could be cumbersome if it added third-party software to the existing workflow. A software package could be employed that would import the values from the .csv file export from FUSION™, enabling PCA analysis of a batched run or individual samples and calculation of the Euclidean distance ratio to determine if the sample results from the run were abnormal compared to the patient’s previous history. However, if the lightweight PCA-based algorithm was added as a feature to existing software that is utilized by HLA laboratories to analyze HLA antibody data, such as HLA FUSION™ (One Lambda; Thermo Fisher Scientific Inc., Waltham, Massachusetts, United States of America), mTilda (HLA Data Systems LLC, Houston, Texas, United States of America), HISTOTRAC™ (System Link, Inc., a part of Thermo Fisher Scientific Inc., Waltham, Massachusetts, United States of America), and / or MATCH IT! ® (Immucor, Inc., now a part of Werfen, Norcross, Georgia, United States of America) among others, it would allow for the automated screening of samples to identify samples with significant variance from their historic antibody pattern and those that are comparable to patient antibody history. This would both simplify analysis for laboratory staff analyzing the testing results and decrease the amount of time necessary to analyze batches of patient samples, likely decreasing sample turnaround time.

[0082] An all too common complication in HLA antibody testing is the over-reactivity or false reactivity observed for some HLA single antigen beads (Dean et al., 2020; Kim et al., 2021). This can be the result of neo-epitope exposure and unspecific binding of serum matrix proteins. As a consequence, some single antigen beads are more prone to false reactivity or over-reactivity compared to other beads in the assays. To evaluate if PCA could provide a method for identifying this increased variance across beads that display a known pattern of over-reactivity, 121 single antigen bead HLA class II sample results that demonstrated overreactive DPAl*02:01-DPB 1*01 :01 and DPA 1*02:02- DPBl*05:01 beads were analyzed. The results showed that the presently disclosed method could distinguish the two overreactive beads from all of the other beads in the HLA class II single antigen bead panel (see Figure 6). This same method was applied to all single antigen bead HLA class I and HLA class II sample results that had been tested by the instant co-inventors from January 2018 to March 2022 that met the sample inclusion criteria. This was done to investigate if there were beads that consistently exhibit over reactivity across multiple lots.

[0083] As disclosed herein, it was determined that for HLA class I, the beads that exhibited the most over-reactivity included A*29:01, A*29:02, and B*15: 12, among others (see Table 3). The most overreactive beads for HLA class II included those for DRB3*02:02, DPAl*02:01-DPB 1*01 :01, DPAl*02.02-DPBl*05:01, DQAl*05:05- DQBl*03:01, DRBl*04:01, and DRBl*04:03, among others (Table 4). These results demonstrated that these beads had a higher variance across the thousands of patient samples tested compared to all of the other beads for the HLA class I and HLA class II single antigen bead assays. This type of analysis is beneficial for the clinical evaluation of patient results. Using the presently disclosed method allows a user to understand which beads have the propensity to demonstrate false reactivity. Once identified, these beads can merit closer scrutiny when they might have an impact on patient care.

[0084] The presently disclosed method can be applied to each laboratory antibody testing database once antibody MFI values have been exported from whatever analysis software is employed. This would provide a customized assessment of which beads display over-reactivity given the laboratory patient population and testing protocol, as certain procedures can reduce bead over-reactivity by using additional washes or treatments.

[0085] The presently disclosed method can also be utilized to assess bead lots. If there are historic beads with a new recombinant or native HLA protein, HLA complex attachment to the beads using a different chemistry or any other production-related procedure change that could potentially impact bead reactivity upon testing, this method could be utilized to assess the bead lot(s) across one or multiple laboratory test sites.

[0086] In identifying differences in patients’ current HLA antibody patterns from historic HLA antibody patterns, only reactivity above 1,000 MFI was thoroughly investigated, since that is the cut-off for reactivity established for the assay and part of the employed inclusion criteria. However, it is well known that real HLA antibodies can be present and represented in these assays with MFI values below 1,000 MFI due to the spreading of antibodies over multiple beads and / or as a result of low concentrations in patient serum. The efficacy of the presently disclosed method on patient samples with only low MFI values (e.g., below 1,000 MFI) could be investigated with further rigor utilizing additional samples with characteristics that would best assess the performance of the presently disclosed method. However, the presently disclosed method are nonetheless expected to yield similar results on samples with real HLA reactivity below 1,000 MFI. It is noted that the presently disclosed proof of concept study utilized only a single HLA class I and HLA class II single antigen bead run, including only a small number of patients. If the process were built and more automated so that it could be time-saving, it would be beneficial to perform a similar study and include many more patients to more stringently assess the sensitivity and specificity of the method on a larger number of patient samples.

[0087] For the implementation of these methods in the clinical laboratory, a package specific for its utility in R or incorporation of this type of method in new or existing HLA antibody analysis software can be employed. This allows for practical testing of a large number of patient samples to further characterize its utility. It also provides valuable data to demonstrate and delineate the time saved in assessing patient samples as a screening mechanism.

[0088] In summary, the presently disclosed results demonstrated the utility of using PCA-derived analysis to facilitate the assessment of LUMINEX® solid phase HLA antibody testing results. Integrating PCA-derived analysis would enable automated screening of patient samples for HLA antibody testing, identifying samples suspected of deviant HLA antibody patterns compared to previous test results. Such deviances from historic patterns could include resurgent DSA, dnDSA, increased antibody reactivity, and prozone activity. The presently disclosed method could also contribute to providing evidence for, and the identification of, over and / or false reactive single antigen beads in both the HLA class I and HLA class II assays.

[0089] In the present study, solid organ transplant patients who exhibited HLA donorspecific antibodies (DSAs) were used to characterize the utility of PCA-derived analysis when compared to a control group of post-transplant and pre-transplant patients. ROC analysis was utilized to determine a potential threshold for the PCA- derived analysis that would indicate a significant change in a patient’s single antigen bead pattern. To evaluate if the algorithm could identify differences in patterns on HLA class I and HLA class II single antigen bead results using the optimized threshold, HLA antibody test results were analyzed using PCA-derived analysis and compared to the clinical results for each patient sample. The PCA-derived algorithm had a sensitivity of 100% (95% CI, 73.54%-100%), a specificity of 75% (95% CI, 56.30%-92.54%), with a PPV of 65% (95% CI, 52.50%-83.90%) and an NPV of 100%, in identifying new reactivity that differed from the patient’s historic HLA antibody pattern. Additionally, PCA-derived analysis was utilized to assess the potential over-reactivity of single antigen beads for both HLA class I and HLA class II antibody panels. This assessment of antibody results identified several beads in both the HLA class I and HLA class II antibody panel which exhibit over reactivity from 2018 to the present time.

[0090] PCA-derived analysis would be ideal to help automatically identify patient samples that have an HLA antibody pattern of reactivity consistent with their history and those which exhibit changes in their antibody patterns which could include donorspecific antibodies, de novo HLA antibodies, and assay interference. A similar method could also be applied to evaluate the over-reactivity of beads in the HLA solid phase assays which would be beneficial for lot comparisons and instructive for transplant centers to better understand which beads are more prone to exhibiting over-reactivity and impact patient care.

[0091] REFERENCES

[0092] All references listed below, as well as all references cited in the instant disclosure, including but not limited to all patents, patent applications and publications thereof, scientific journal articles, and database entries (e.g., GENBANK® and UniProt biosequence database entries and all annotations available therein) are incorporated herein by reference in their entireties to the extent that they supplement, explain, provide a background for, or teach methodology, techniques, and / or compositions employed herein.

[0093] Amico et al. (2009) Clinical relevance of pretransplant donor-specific HLA antibodies detected by single-antigen flow-beads. Transplantation 87: 1681-1688.

[0094] Bray et al. (2011) Chapter 12 - Clinical Cytometry and Progress in HLA Antibody Detection. In: Darzynkiewicz et al. (eds.) Methods in Cell Biology. Academic Press, 103:285-310.

[0095] Dean et al. (2020) Identification of a recurrent pattern of false-positivity by Luminex HLAMHC class I single antigen bead testing. Human Immunology 81(2-3):73- 78.

[0096] Elmore & Richman (2001) Euclidean Distance as a Similarity Metric for Principal Component Analysis. Monthly Weather Review 129:540-549.

[0097] Gandhi et al. (2013) Lot-to-lot variability in HLA antibody screening using a multiplexed bead-based assay. Transfusion 53: 1940-1947.

[0098] Garate-Escamila et al. (2020) Classification models for heart disease prediction using feature selection and PCA. Informatics in Medicine Unlocked 19: 100330.

[0099] Kherif & Latypova (2020) Chapter 12 - Principal component analysis. In: Mechelli & Vieira (eds). Machine Learning: Methods and Applications to Brain Disorders. Academic Press 2020:209-225.

[0100] Kim et al. (2021) False-positive reactivity of anti-human leukocyte antigen antibodies detected using the single-antigen bead assay. Human Immunology 82:409-413.

[0101] Kuncoro & Sugiharti (2021) Improved accuracy of recognizing of low-quality face images using two directional matrix in 2D-PCA algorithms and euclidean distance. J Phys: Conf Ser 1918:042142.

[0102] Mamode et al. (2022) European Guideline for the Management of Kidney Transplant Patients With HLA Antibodies: By the European Society for Organ Transplantation Working Group. Transpl Int 35: 10511.

[0103] Novembre et al. (2008) Genes mirror geography within Europe. Nature 456:98-101.

[0104] Patel & Terasaki (1969) Significance of the Positive Crossmatch Test in Kidney Transplantation. New England Journal of Medicine 280:735-739.

[0105] Sari et al. (2018) Iris recognition based on distance similarity and PCA. AIP Conf Proc 1977:020044.

[0106] Verleden et al. (2017) Donor-specific and -nonspecific HLA antibodies and outcome post lung transplantation. European Respiratory Journal 50: 1701248.

[0107] Vittoraki et al. (2020) Patterns of 1,748 Unique Human Alloimmune Responses Seen by Simple Machine Learning Algorithms. Frontiers in Immunology Front Immunol 11 : 1667.

[0108] Vittoraki et al. (2021) Hidden Patterns of Anti -HLA Class I Alloreactivity Revealed Through Machine Learning. Front Immunol 12: 670956.

[0109] Wehmeier et al. (2020) Caveats of HLA antibody detection by solid-phase assays. Transplant Int 33: 18-29.

[0110] Weimer & Newhall (2019) Development of data-driven models for the flow cytometric crossmatch. Human Immunology 80:983-989.

[0111] Wiebe et al. (2012) Evolution and clinical pathologic correlations of de novo donorspecific HLA antibody post kidney transplant. American Journal of Transplantation 12: 1157-1167. Yanagida et al. (2011) Impact of virtual cross match on waiting times for heart transplantation. Annals of Thoracic Surgery 92:2104-2111.

[0112] Yu et al. (2022) Current HLA testing recommendations to support HCT. Human Immunology 83:665-673.

[0113] Zhi & Liu (2019) Face recognition based on genetic algorithm. Journal of Visual Communication and Image Representation 58:495-502.

[0114] It will be understood that various details of the subject matter described herein may be changed without departing from the scope of the subject matter described herein. Furthermore, the foregoing description is for the purpose of illustration only, and not for the purpose of limitation, as the subject matter described herein is defined by the claims as set forth hereinafter.

Claims

CLAIMSWhat is claimed is:

1. A method for analyzing human leukocyte antigen (HLA) test results to identify clinically relevant changes in test patterns or evaluate HLA testing, the method comprising:(a) receiving, as input, HLA test results;(b) using principal component analysis (PCA) to analyze the HLA test results to detect a clinically relevant change for a subject or evaluate an aspect of HLA testing; and(c) outputting results of the analysis.

2. The method of claim 1, wherein the HLA test data includes historical HLA test data for the subject.

3. The method of claim 2, wherein the historical HLA test data includes HLA test data before and after a donor specific antibody (DSA) was identified in samples for the subject.

4. The method of claim 1, wherein using PCA to analyze the HLA test results includes using PCA to identify antigen bead over-reactivity.

5. The method of claim 1, wherein using PCA to analyze the HLA test results includes using PCA to evaluate differences in reactivity between antigen bead lots.

6. A method for evaluating performance of a second set of HLA testing reagents as compared to a first set of HLA testing reagents, the method comprising:(a) receiving, as input, HLA test results from the second set of HLA testing reagents;(b) using principal component analysis (PCA) to analyze the HLA test results from the second set of HLA testing reagents; and(c) comparing the HLA test results from the second set of HLA testing reagents to equivalent HLA test results from the first set of HLA testing reagents, wherein a difference between the HLA test results from the second set of HLA testing reagents and the HLA test results from the first set of HLA testing reagents is indicative of a difference in performance of the second set of HLA testing reagents as compared to the first set of HLA testing reagents; and optionally(d) outputting results of the analysis.

7. The method of claim 6, wherein the comparing is indicative of the second set of HLA testing reagents performing less effectively than the first set of HLA testing reagents with respect to identifying presence of donor specific antibodies (DS As) in a biological sample isolated from a subject.

8. The method of claim 6 or claim 7, wherein the first and second sets of HLA reagents comprise HLA class I beads, HLA class II beads, or a combination thereof.

9. A method for assessing HLA antibody testing, the method comprising:(a) receiving, as input, a first set of HLA test results;(b) using principal component analysis (PCA) to analyze the first set of HLA test results; and(c) comparing the first set of HLA test results to a second set of HLA testing results, wherein the first and second set of HLA testing results are derived from testing of a biological sample with a first and second set of HLA testing reagents, respectively, under equivalent conditions; and optionally(d) outputting results of the analysis, wherein a difference between the first and second sets of HLA testing is indicative of a difference in performance of the first and second sets of HLA testing reagents.

10. The method of claim 9, wherein the comparing is indicative of the second set of HLA testing reagents performing less effectively than the first set of HLA testing reagents with respect to identifying presence of donor specific antibodies (DSAs) in the biological sample.

11. The method of claim 9 or claim 10, wherein the first and second sets of HLA reagents comprise HLA class I beads, HLA class II beads, or a combination thereof.

12. The method of any one of the preceding claims, wherein the HLA test results are generated by HLA testing of a biological sample isolated from a subject, and further wherein the subject is a pre-transplant subject or a post-transplant subject.

13. The method of claim 12, wherein the HLA test results comprise HLA test datacomprising HLA test data related to the presence of donor specific antibodies (DS As) in the subject, optionally wherein the presence of DSAs was identified in samples for the subject before and / or after a solid organ or a hematopoietic transplant into the subject.

14. A non-transitory computer readable medium having stored thereon executable instructions that when executed by a processor of a computer control the computer to perform the method as claimed in any of the preceding claims.

Images