Quality control metrics for methylation assays

Abstract

The presently described techniques utilize detection p-value metrics as part of determining sample quality in the context of a methylation assay. In these approaches, the detection p-value may be understood to be the probability of observing a signal at least as extreme as the measured or detected signal. Such a detection p-value is used to assess quality of individual probes. Based on the detection p-values for a given set of probes, the quality of a given sample may be assessed as a detection rate, which may be understood to be the proportion of probes passing a p-value detection threshold.

Description

QUALITY CONTROL METRICS FOR METHYLATION ASSAYSTECHNICAL FIELD

[0001] The disclosed technology relates generally to techniques and metrics for assessing the quality of a sample used in an array -based methylation procedure. In particular, the disclosed techniques relate to detection p-value methods and to differing permutations of factors that may be used in assessing the quality of a sample.BACKGROUND

[0002] The subject matter discussed in this section should not be assumed to be prior art merely as a result of its mention in this section. Similarly, a problem mentioned in this section or associated with the subject matter provided as background should not be assumed to have been previously recognized in the prior art. The subject matter in this section merely represents different approaches, which in and of themselves can also correspond to implementations of the claimed technology.

[0003] DNA methylation plays a role in the regulation of gene expression. Methylation assays, such as those performed using bead arrays, may be used to perform accurate, quantitative assessment of DNA methylation at the single-CpG-site level, thereby allowing the acquisition of data that may be relevant to regulation of expression of a gene or genes in a subject. This may facilitate investigation into epigenetic changes in various health and / or diagnostic contexts. In particular, bead-based methylation assays may provide high- throughput capabilities in conjunction with broad coverage to detect epigenetic modifications. Correspondingly, such methylation assays may be useful in context epigenome studies, such as large-scale epigenome-wide association studies (EWAS).

[0004] In general, such bead-based methylation assays employ beads having target-specific probes configured to interrogate individual CpG sites within a given sample. In certain implementations an assay may employ two probes (one for a methylated state and one for anunmethylated state) per CpG locus, while in other implementations a single probe (capable of measuring both an unmethylated and a methylated state) per CpG locus may be employed.

[0005] While such assay designs and tools are useful, they are subject to various constraints. For example, various tertiary analyses (e.g., methylation-based risk scores, differential methylation metrics or scores, and so forth) may necessitate large sample sizes. Correspondingly, determination of the quality of a given sample may be a factor evaluating whether metrics based on the methylation assay results for that sample are reliable.

[0006] However existing quality control techniques for assessing a sample are either manual (and thus subject to subjective assessment as well as being time consuming) or are otherwise low throughput. Techniques are therefore needed for rapidly and reliably assessing sample quality for bead-based methylation assays.BRIEF DESCRIPTION

[0007] The presently described techniques utilize detection p-value metrics as part of determining sample quality in the context of a methylation assay. In these approaches, the detection p-value may be understood to be the probability of observing the measured signal under the null hypothesis in a one-sided test. Such a detection p-value may be used to assess the quality of individual probes (e.g., negative control probes). Based on the detection p- values for a given set of probes, the quality of a given sample may be assessed as a detection rate, which may be understood to be the proportion of probes passing a p-value detection threshold.

[0008] Within a bead array design, control probes may be provided as bead array controls reporter(s) (BACR), which yield different metrics across various quality-related categories. While such metrics may individually provide limited information about sample quality, the presently described techniques utilize novel combinations (e.g., permutations) of metric data in assessing sample quality. In particular, for control probe data for a given sample, permutations of metrics for distribution type (e.g., empirical cumulative distribution function(ECDF) or normal), channel-specific distribution (e.g., red and green channels in view of Type 1 and Type 2 probes utilized), and logarithmic transformation (e g., data log transformed or not log transformed) were analyzed using data sets known to have high-failure rates. Based on the analyses, the combinations of certain BACR metrics were determined to be suitable for assessing quality of methylation array sample data.

[0009] In one embodiment, the present disclosure provides a processor-based method for assessing sample quality. In accordance with this method, a sample is processed using a methylation screening assay comprising a plurality of negative control probes. The methylation screening assay generates an output for the sample that comprises a plurality of metrics based on the hybridization of the negative control probes to the sample. Detection p- values are calculated for probes of the methylation screening assay based on the plurality of metrics, wherein the detection p-values are calculated based on at least: a normal distribution and a logarithmic transformation of signal data. A detection rate is calculated based on the detection p-values. The detection rate is output as a sample quality metric. In further embodiments processor-executable code for performing such actions may be stored on one or more non-transitory, computer-readable media or as part of a processor-based system.

[0010] In one embodiment, the present disclosure provides a processor-based method for assessing sample quality. In accordance with this method, a sample is processed using a methylation screening assay comprising a plurality of negative control probes. The methylation screening assay generates an output for the sample that comprises a plurality of metrics based on the hybridization of the negative control probes to the sample. Detection p- values are calculated for probes of the methylation screening assay based on the plurality of metrics, wherein the detection p-values are calculated based on at least: a normal distribution and channel-specific distributions. A detection rate is calculated based on the detection p- values. The detection rate is output as a sample quality metric. In further embodiments processor-executable code for performing such actions may be stored on one or more non- transitory, computer-readable media or as part of a processor-based system.BRIEF DESCRIPTION OF THE DRAWINGS

[0011] These and other features, aspects, and advantages of the disclosed embodiments will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:

[0012] FIG. 1 depicts a stylized representation of a methylation assay workflow, in accordance with aspects of the present techniques;

[0013] FIG. 2 illustrates graphically the determination of a detection p-value, in accordance with aspects of the present techniques;

[0014] FIG. 3 illustrates schematically a computational pipeline of a methylation assay and input and outputs of such a pipeline, in accordance with aspects of the present techniques;

[0015] FIG. 4 depicts an example of a sample output of the pipeline of FIG. 3, in accordance with aspects of the present techniques;

[0016] FIG. 5 depicts a further example of a sample output of the pipeline of FIG. 3 depicting sample quality control metrics, in accordance with aspects of the present techniques;

[0017] FIG. 6 depicts plots depicting the samples passing all BACR metrics (left) and the samples failing at least one BACR metric (right), in accordance with aspects of the present techniques;

[0018] FIG. 7 depicts additional plots depicting the samples passing all BACR metrics (left) and the samples failing at least one BACR metric (right), in accordance with aspects of the present techniques;

[0019] FIGS. 8A and 8B collectively depict, in tabular form, correlations of BACR metrics with detection rate for data sets analyzed using the EPIC chip, in accordance with aspects of the present techniques;

[0020] FIGS. 9A and 9B collectively depict, in tabular form, correlations of BACR metrics with detection rate for data sets analyzed using the MSA chip, in accordance with aspects of the present techniques;

[0021] FIG. 10 depicts a plot of paired samples of different sample size, in accordance with aspects of the present techniques;

[0022] FIG. 11 is a plot of the Pearson correlation of Betas versus DNA input, with data points representing samples that passed quality control (value 1) and those that failed quality control (value 0), in accordance with aspects of the present techniques;

[0023] FIG. 12 depicts a Venn diagram depicting the overlaps of failed sample determinations made by manual reviewers and the presently described automated techniques, in accordance with aspects of the present techniques;

[0024] FIG. 13 depicts an AUC analysis of the results of the manual reviewers and the presently described automated techniques, in accordance with aspects of the present techniques; and

[0025] FIG. 14 depicts components of a methylation assay analysis system, in accordance with aspects of the present techniques.DETAILED DESCRIPTION

[0026] The following discussion is presented to enable any person skilled in the art to make and use the technology disclosed, and is provided in the context of a particular application and its requirements. Various modifications to the disclosed implementations will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other implementations and applications without departing from the spirit and scope of the technology disclosed. Thus, the technology disclosed is not intended to be limited to the implementations shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0027] DNA methylation plays a role in regulating gene expression. Accordingly, assessments of methylation are a useful tool in evaluating genomic (e.g., genome- wide association studies (GWAS)) and epigenomic (e.g., epigenome-wide association studies (EWAS)) data. In particular, such assessments may be useful in determining the presence or absence of cancer or other gene regulation irregularities. In certain types of assays, such assessments may quantify DNA cytosine modifications in the form of methylation, which is in turn associated with the control of a given gene’s expression. Tools for assessing methylation, however, typically require large sample sizes and are susceptible to artifacts or poor signal quality associated with deletions present within the sequence in question. Correspondingly, assessing the quality of a given sample for which a methylation metric or measurement has been calculated may be valuable in determining whether results for that given sample or assay are accurate and / or can be relied upon. In particular, determination of whether a given sample meets a threshold quality level (i.e., is “good enough”) is one problem associated with current tertiary analyses (e.g., methylation risk scores, differential methylation analyses, and so forth) of methylation assay data (e.g., bead-based assays) due to the sensitivity of such assays to sample quality. In practice such a quality control metric may be a direct or indirect indication of sample quality (in terms of insertions, deletions, or other sequence inconsistencies) and / or sample representativeness. With respect to secondary or tertiary analyses for which such a quality metric may be relevant (such as risk scores, differential comparisons, and so forth), the significance of the quality control metric may be displayed with the analyses results (such as a displayed numeric quality control metric) and / or may be incorporated into the display of such analyses results, such as in the form of a color-coded indicator or other visual indicator, confidence bars or bands, and so forth. Alternatively, the quality control metric may not itself be displayed, but such color coding or visual indicators may still be displayed to indicate the trustworthiness of analysis results in view of sample quality.

[0028] With the preceding in mind, the described techniques utilize detection p-value metrics as part of determining sample quality in the context of a methylation assay. In these approaches, the detection p-value may be understood to be the probability of observing a signalat least as extreme as the measured or detected signal under the null hypothesis in a one-sided test. Such a detection p-value is used to assess quality of individual probes (e.g., negative control probes). Based on the detection p-values for a given set of probes, the quality of a given sample may be assessed as a detection rate, which may be understood to be the proportion of probes passing a p-value detection threshold.

[0029] As used herein, the negative control probes employed correspond to actual DNA sequences for which nucleotide positions have been shuffled, thereby creating a probe sequence that is expected to have low hybridization and, in this manner, effectively provide a “background” signal for comparison or removal. Within a bead array design, these may be provided as bead array controls reporter(s) (BACR), which in accordance with present designs yield different metrics (e.g., 21 metrics) across different categories (e.g., 8 categories, such as, but not limited to, restoration, staining, extension, and so forth).

[0030] While such metrics may individually provide limited information about sample quality, the presently described techniques utilize novel combinations of metric data or characteristics in assessing sample quality. In particular, for negative control probe data for a given sample, permutations of metrics for distribution type (e.g., empirical cumulative distribution function (ECDF) or normal), channel-specific distribution (e.g., red and green channels in view of Type 1 and Type 2 probes utilized), and logarithmic transformation (e.g., data log transformed or not log transformed) were analyzed using data sets known to have high-failure rates. Based on the analyses, different combinations of BACR metrics were evaluated for their suitability in assessing quality of methylation bead-array sample data.

[0031] With the preceding in mind, and turning to the figures, FIG. 1 depicts a stylized representation of stages (e.g., a workflow) and / or aspects of a methylation assay as described herein. In the depicted example, the configuration and / or structure of the bead chip and the corresponding content (represented by reference number 100) may comprise various factors, such as but not limited to: bead chip formats (e.g., 48 and 24 EX or other high throughput formats); content (e.g., Methylation Screening Array, EPIC-Enhancer Array, and so forth); control probes present on the array; and / or manifest generation details for the array. In asecond aspect, the characteristics or details related to the automation (represented here by reference number 104) may comprise various factors, such as but not limited to: automation of a methylation workflow (e.g., from bisulfite conversion to intensity data); automation with respect to a laboratory information management system (LIMS) and / or lab automation software solution; and / or automation with respect to bisulfite conversion lightning kits. In a third aspect, the characteristics or details related to the assay reagents (represented here by reference number 108) may comprise various factors, such as but not limited to: the sample input requirement (e.g., 25 ng, 35 ng, 50 ng, 65 ng, and so forth); appropriate assay reagents, and / or sample type. In a fourth aspect, the characteristics or details related to the analyses (e.g., secondary or tertiary analyses, represented here by reference number 112) may comprise various factors, such as but not limited to: basespace workflow; Beta / M values; bead array control calling; computation of detection p-values, and / or static plots and output files. As discussed herein, the presently described techniques relate primarily to the analysis performed at the analysis stage.

[0032] As used herein, the phrase “detection p-values” may be understood to correspond to the probability of observing a signal at least as extreme as the observed data under the null hypothesis (one-sided). This is illustrated graphically in FIG. 2, where the detection p-value is illustrated as corresponding to the shaded area under the distribution and to the right of the observed data point. In practice, the detection p-values may be used to perform quality control (QC) of individual probes (e.g., each probe) used in the methylation array. A threshold of 0.05 is typically employed to distinguish observed values that correspond to expectations and those that significantly differ from expectations. In practice, the detection p-values may be calculated using “negative control probes”, which may be understood to be actual DNA sequences for which nucleotide positions have been shuffled, thereby creating a probe sequence that is expected to have low hybridization and, in this manner, effectively provide a “background” signal for comparison purposes.

[0033] While “detection p-values” relate to quality control at the probe level, “detection rate” corresponds to a quality control metric at the sample level. In this context, “detection rate”may be understood to be the proportion of probes passing a p-value detection threshold, as described above. In practice, as noted here, the detection rate may be employed as quality control for a sample and, typically, may have a corresponding threshold to assess statistical significance, which may vary based on factors including, but not limited to, type of sample, DNA input amount, and so forth.

[0034] In further aspects, in practice a methylation array may employ metrics based on or otherwise derived using control probes (e.g., negative control probes) as discussed herein. Such bead array controls reporter(s) (BACR) may provide quality control information that can be used to basic sample quality control, but which may fall short of a general metric for assessing sample quality. Such BACR probes may be specifically designed so as to measure success or sufficiency of certain parts of the assay workflow (as discussed with respect to FIG. 1) and such analyses may summarize individual control probe values for easy interpretation. An example of such BACR probes and their corresponding workflow correspondence is shown in Table 1.Table 1

[0035] With the preceding in mind, and turning to FIG. 3, an example methylation caller overview is depicted in terms of inputs 200 to a computational pipeline, the computational pipeline 204 (including operations performed by the pipeline), and outputs 208 of the pipeline. In this example of input 200 and outputs 208, required inputs and outputs of the example pipeline 204 are shown in shaded boxes and optional inputs are shown using broken lines. By way of example, inputs 200 comprise an intensity data (ID AT) directory, a manifest, and an output directory. Optional inputs 200 in this example include a sample sheet and custom run parameters. Operations of the pipeline 204 include, but are not limited to, converting the manifest format, generating and / or formatting the sample sheet, computing beta values, p- values (e.g., detection p-values), and / or BACR metrics, and computing a sample quality control metric, as discussed herein. These pipeline operations are reflected in the outputs 208 of this example pipeline. In particular, such outputs 208 may include, but are not limited to, beta values, M-values, p-values (e.g., detection p-values), a sample quality control report, and / or sample quality control plots. As used herein, Beta (ft) values may be understood to correspond to:where M represents a hybridization signal from a methylated version of a cytosine nucleotide, U represents the hybridization signal from an unmethylated version of the cytosine nucleotide, and a is an offset. In practice, the offset a may be omitted in certain contexts, in which case equation (1) may be effectively reduced to:As discussed herein, a ft value should preferably be reported only for actual (i.e., true) signals corresponding to a hybridization event, as opposed to background fluorescence (as may be assessed using negative control probes). As may be appreciated, the probe-level detection p- values, as discussed herein, may be used in assessing whether a signal is deemed to be actual (i.e., true) or not.

[0036] Turning to FIGS. 4 and 5, sample outputs 208 of the computational pipeline 204 are illustrated. In this example FIG. 4 displays sample specific outputs 208, e.g., beta ( / >) values, M values (i.e., methylation values or score), and p-values. Turning to FIG. 5, additional outputs 208 may include quality control outputs for each sample, which in this example include, but are not limited to: the BACR metrics, summaries of signals (here depicted as two channels (red and green) having raw and normalized signals), scan time, detection rate, a pass / fail quality control assessment, and a listing of failed quality control metrics. In addition, graphical outputs, such as to identify poor performing samples, may be provided among the outputs 208. Such graphical outputs may include, but are not limited to, principal component analyses (PCA) of beta (fl) values passing a p-value threshold, distributions of detection rate by chip, and so forth.

[0037] With the preceding in mind, the presently described techniques relate to the identification and use of new methodologies for calculating detection p-values using negative control probes (i.e., shuffled sequences added to an array that are expected to have low hybridization to DNA), as shown in the “Probes used for null” column of Table 2. In this example, only one reference technique, pOOBAH, did not use the same negative control probe scheme, instead employing an out-of-band probe scheme. Other relevant parameters that were reviewed included the distribution type of the data or signals (e.g., ECFD or normal distributions), whether channel-specific distributions were employed (e.g., red and green channel specific distributions or a single combined distribution), and whether signal data was or was not logarithmically transformed. Various techniques and permutations of these parameters (e.g., distribution type, channel-specific (e.g., red and green channels) distributions; and logarithmic data transformation) were analyzed to identify combinations yielding suitable and useful detection p-values. Turning to Table 2, the permutations of analyzed parameters are provided.Table 2In accordance with this table, the permutations included: (1) detectionPnegEcdf (i.e., negative control probes, an ECDF distribution, no channel specific distributions, and no logarithmic transformation of the signal data); (2) detectionPnegNormGS (i.e., negative control probes, a normal distribution, no channel specific distributions, and no logarithmic transformation of the signal data); (3) detectionPnegNorm2 (i.e., negative control probes, a normal distribution, channel specific distributions, and no logarithmic transformation of the signal data); (4) detectionPnegLM (i.e., negative control probes, a normal distribution, no channel specific distributions, and logarithmic transformation of the signal data); and (5) detectionPnegLM2 (i.e., negative control probes, a normal distribution, channel specific distributions, and logarithmic transformation of the signal data).

[0038] To assess the use of the new detection p-values and the permutations of factors noted above, data sets were evaluated that were known to have high sample failure rates. The data sets and the bead assay chip (EPIC or MSA) used in their analysis are illustrated in Table 3.Table 3

[0039] Turning to FIG. 6, plots of the Table 3 chips and data sets corresponding to the EPIC chip and the CFDNA (i.e., genomic and cell-free DNA) data set are shown for the six detection p-value techniques listed in Table 2. In FIG. 6, the left-most plot depicts the samples passing all BACR metrics in a plot of threshold (x-axis) versus detection rate (y-axis). The right-most plot depicts the samples failing at least one BACR metric along the same axes. As visually illustrated, the negative control probe based techniques are more similar to each other when plotted than the pOOBAH ( -value with out-of-band (OOB) array hybridization) technique. In particular, the curves associated with the pOOBAH technique for “samples passing” and “samples failing” tend to be more similar to one another than the corresponding plots for the control probe based techniques. In addition, for a typical p-value threshold of 0.05, the pOOBAH technique had a significantly lower detection rate than the control probe based techniques. FIG. 7 shows similar plots for the MSA chip and the ILS-NIEHS data set.

[0040] Turning to FIGS. 8A and 8B, these figures collectively depict correlations of BACR metrics with detection rate for the CFDNA and EPICvl data sets analyzed using the EPIC chip. The left-most table shows, for the CFDNA data set, correlations of BACR metrics (vertical axis) and detection p-value calculation technique (horizontal axis). The right-mosttable is laid out similarly, but is based on the EPICvl data set. As shown in this correlation data, the detectionPnegLM2 (i.e., (1) control probes used for testing the null hypothesis; (2) normal distribution type; (3) channel-specific (i.e., red channel, green channel) distributions; and (4) logarithmic data transformation) and the detectionPnegLM (i.e., (1) control probes used for testing the null hypothesis; (2) normal distribution type; (3) no channel-specific (i.e., red channel, green channel) distributions; and (4) logarithmic data transformation) techniques for calculating detection p-values exhibited the highest average absolute correlations.

[0041] Turning to FIGS. 9A and 9B, these figures collectively depict similar tables of correlations of BACR metrics with detection rate for the ILS-NIEHS and CHOP-ILS-IDATs data set analyzed using the MSA chip. The left-most table shows, for the ILS-NIEHS data, correlations of BACR metrics (vertical axis) and detection p-value calculation technique (horizontal axis). The right-most table is laid out similarly, but is based on the CHOP-ILS- IDATs data set. As shown in this correlation data, the pOOBAH and detectionPnegEcdf (i.e., (1) control probes used for testing the null hypothesis; (2) ECDF distribution type; (3) no channel-specific distributions; and (4) no logarithmic data transformation) techniques for calculating detection p-values exhibited the highest average absolute correlations.

[0042] In practice, using such correlation data and area-under-curve (AUC) plots and analyses, a user may predict BACR pass status using detection rates calculated for the detection p-values calculated using the presently described techniques. Similarly, for each of the detection p-value techniques described herein, optimal p-value thresholds may be determined, such as by plotting pass and fail BACR counts versus pass and fail thresholds, wherein the corresponding detection rate corresponds to an optimal threshold.

[0043] Turning to Table 4, a summary of AUC results for the above references data sets, methylation assay chips, and detection p-value calculation techniques are provided.Table 4As laid out, the first two data sets (EPICvl and CFDNA) are those analyzed using the Epic methylation assay chip and are averaged together in the “Avg EPIC” column. The next two data sets (ILS-NIEHS and CHOP -ILS) are those analyzed using the MSA methylation assay chip and are averaged together in the “Avg MSA” column. All data set results for both chips are averaged in the “Avg all” column.

[0044] With the preceding in mind, paired samples at low input volume (50 ng) and high input volume (250 ng) were analyzed for sample quality using the preceding techniques and detection p-value techniques. Results are illustrated in FIGS. 10 and 11 for the 240 pairs of samples, with FIG. 10 depicting a plot of density versus concentration of DNA. FIG. 11 is a plot of the Pearson correlation of Betas versus DNA input for the smaller of two replicates, with data points representing samples that passed quality control (value 1) and those that failed quality control (value 0). As shown, only 7 samples failed BACR based quality control and all were the smaller sample size (i.e., 50 ng). The failed samples tended to have a lower Beta correlation with the higher DNA input sample (i.e., 250 ng). With this in mind, in accordance with the present techniques samples with poor quality with respect to methylation assay analysis can be accurately identified.

[0045] In terms of establishing the suitability of the present sample quality control assessment techniques, a study was performed comparing the use of the detectionPnegLM detection p-value calculation technique with the sample quality determinations two manual reviewers using the CHOPS-ILS dataset. Results are illustrated in FIGS. 12 and 13. FIG. 12 depicts a Venn diagram of failed samples as determined by the two manual reviewers and the present automated quality control techniques. FIG. 13 depicts an AUC analysis of the results. As may be seen from these results, the presently described automated sample quality control results are more concordant with detection rates than manual quality control analyses.

[0046] With the preceding discussion in mind, it may be appreciated that the presently described sample quality control metrics may be useful in the context of validating (or otherwise providing an indication of reliability or confidence) methylation assay results, such as tertiary analysis results in the form of methylation-based risk scores, differential methylation metrics or scores, and so forth. In particular, such quality control metrics may help confirm that a sample used in a methylation assay was of sufficient size or sequence quality to justify reliance on the methylation assay results. In practice, a practitioner may select a suitable permutation of the parameters described herein based on their prior experience and / or based upon known characteristics of the sample. Alternatively, a quality control metric based on a single permutation of the parameters as described herein may be used as the preferred technique for generating quality control metrics (e.g., detection p-values, detection rates, and so forth) for a given practitioner or institution. By way of example, in certain embodiments one or more of the detectionPnegLM, detectionPnegLM2, or detectionPnegNorm2 techniques may be selected for the calculation of detection p-values as discussed herein.

[0047] As discussed herein, assessing the quality of a given sample for which a methylation metric or measurement has been calculated may be useful in assessing whether results for the given sample or assay are accurate and / or can be relied upon. In particular, determination of whether a given sample meets a threshold quality level (i.e., is “good enough”) may be useful in the context of tertiary analyses (e.g., methylation risk scores, differential methylation analyses, and so forth) or other downstream analyses of methylation assay data due to the sensitivity of such analyses to sample quality. In particular, such a quality control metric maybe a direct or indirect indication of sample size and / or sample quality (in terms of insertions, deletions, or other sequence inconsistencies). With respect to secondary or tertiary analyses for which such a quality metric may be relevant (such as risk scores, differential comparisons, and so forth), the statistical significance of the quality control metric may be displayed with the analyses results (such as a displayed numeric quality control metric) and / or may be incorporated into the display of such analyses results, such as in the form of a color-coded indicator or other visual indicator, confidence bars or bands, and so forth. Alternatively, the quality control metric may not itself be displayed, but such color coding or visual indicators may still be displayed to indicate the trustworthiness and / or validity of analysis results in view of the determined sample quality.

[0048] FIG. 14 is a schematic diagram of an assay and assay reading device 500 that may be used in conjunction with the disclosed embodiments for acquiring methylation assay data as generally discussed herein.

[0049] In the depicted embodiment, the methylation assay system 500 includes a separate sample processing device 502 and an associated computer 504. However, as noted, these may be implemented as a single device. Further, the associated computer 504 may be local to or networked or otherwise in communication with the sample processing device 502. In the depicted embodiment, the biological sample may be loaded into the sample processing device 502 on a sample substrate 510, e.g., a methylation bead array or chip, that is processed to generate methylation data. The chip or other assay substrate readout device may be based upon any suitable technology.

[0050] The readout device 512 may be under processor control, e.g., via a processor 514, and the sample processing device 502 may also include I / O controls 516, an internal bus 518, nonvolatile memory 520, RAM 522 and any other memory structure such that the memory is capable of storing executable instructions, and other suitable hardware components. Further, the associated computer 504, if separate, may also include a processor 524, VO controls 526, communications circuity 527, and a memory architecture including RAM 528 and non-volatile memory 530, such that the memory architecture is capable of storing executable instructions532. The hardware components may be linked by an internal bus, which may also link to the display 534. In embodiments in which the methylation assay system 500 is implemented as an all-in-one device, certain redundant hardware elements may be eliminated.

[0051] This written description uses examples to enable any person skilled in the art to practice the disclosed embodiments, including making and using any devices or systems and performing any incorporated methods. The patentable scope is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal languages of the claims.

Claims

CLAIMSWhat is claimed is:

1. A method for assessing sample quality comprising: processing a sample using a methylation screening assay comprising a plurality of negative control probes, wherein the methylation screening assay generates an output for the sample that comprises a plurality of metrics based on the hybridization of the negative control probes to the sample; calculating detection p-values for probes of the methylation screening assay based on the plurality of metrics, wherein the detection p-values are calculated based on at least: a normal distribution; and a logarithmic transformation of signal data; calculating a detection rate based on the detection p-values; and outputting the detection rate as a sample quality metric.

2. The method of claim 1, wherein the methylation screening assay is a bead array assay.

3. The method of claim 1, wherein the sample is a nucleic acid sample.

4. The method of claim 3, wherein the nucleic acid sample is a DNA sample.

5. The method of claim 1, wherein the detection p-values are calculated without using channel-specific distributions.

6. The method of claim 1, wherein the detection p-values are calculated using channelspecific distributions.

7. The method of claim 1, wherein different categories of negative control probes are configured to measure different respective portions of a workflow of the methylation screening assay.

8. The method of claim 1, wherein the detection p-values correspond to the probability of observing a signal at least as extreme as a measured signal.

9. The method of claim 1, wherein the detection rate corresponds to the proportion of probes of the methylation screening assay passing a p-value detection threshold.

10. The method of claim 1, wherein the negative control probes comprise shuffled sequences expected to have low hybridization to the sample.

11. The method of claim 1 , wherein the sample quality metric identifies samples having accuracy below a sample quality threshold, wherein:wherein M represents a hybridization signal from a methylated version of a cytosine nucleotide, U represents the hybridization signal from an unmethylated version of the cytosine nucleotide, and a is an offset.

12. A method for assessing sample quality comprising: processing a sample using a methylation screening assay comprising a plurality of negative control probes, wherein the methylation screening assay generates an output for the sample that comprises a plurality of metrics based on the hybridization of the negative control probes to the sample; calculating detection p-values for probes of the methylation screening assay based on the plurality of metrics, wherein the detection p-value are calculated based on:a normal distribution; and channel-specific distributions; calculating a detection rate based on the detection p-values; and outputting the detection rate as a sample quality metric.

13. The method of claim 12, wherein the methylation screening assay is a bead array assay.

14. The method of claim 12, wherein the detection p-values are calculated without performing a logarithmic transformation of signal data.

15. The method of claim 12, wherein different categories of negative control probes are configured to measure different respective portions of a workflow of the methylation screening assay.

16. The method of claim 12, wherein the detection p-values correspond to the probability of observing a signal at least as extreme as a measured signal.

17. The method of claim 12, wherein the detection rate corresponds to the proportion of probes of the methylation screening assay passing a p-value detection threshold.

18. The method of claim 12, wherein the negative control probes comprise shuffled sequences expected to have low hybridization to the sample.

19. The method of claim 12, wherein the sample quality metric identifies samples having accuracy below a sample quality threshold, wherein:wherein M represents a hybridization signal from a methylated version of a cytosine nucleotide, U represents the hybridization signal from an unmethylated version of the cytosine nucleotide, and a is an offset.

20. The method of claim 12, wherein the channel-specific distributions comprise a red channel distribution and a green channel distribution.

Images