WO2017205251A1

WO2017205251A1 - Attribute sieving and profiling with sample enrichment by optimized pooling

Info

Publication number: WO2017205251A1
Application number: PCT/US2017/033761
Authority: WO
Inventors: Michael Seul; Ghazala Hashmi
Original assignee: Bioinventors & Entrepreneurs Network, Llc
Priority date: 2016-05-25
Filing date: 2017-05-22
Publication date: 2017-11-30
Also published as: CA3029747A1; EP3465499A4; EP3465499B1; US10208347B2; US20170342489A1; EP3465499A1

Abstract

A process of identifying a plurality of biological samples having particular desirable attributes by testing pooled samples and selecting, for intended uses such as transfusion, or for subsequent analysis that is thereby enriched for such samples, pooled samples which have, or may have, said desirable attributes.. The preferred number of samples per pool "d" is determined by setting d as an integer which produces the maximum or a value near the maximum of the product of: d times the expected number of unambiguous sample pools, where a sample pool is unambiguous if all of the samples have the desirable attributes, and is otherwise ambiguous if at least one sample has the desirable attributes. The value selected as d can be greater than the maximum product above, so as to enlarge the total number of samples assayed in determining the desirable attributes.

Description

Attribute Sieving and Profiling with Sample Enrichment by Optimized Pooling Background

Many clinical applications call for stratification of patients by molecular (and/or other) attributes. For example, to develop or administer personalized therapies, patients may be selected for clinical trials or drug development programs in accordance with molecular attribute profiles that provide a differential diagnosis, such as for many cancers, or indicate the efficacy and/or safety of therapy (see e.g. Doehner2010, Kurose2012).

Further, many patients with cytopenias require regular transfusion. For example, patients with anemia caused by renal disease, or by hematologic disorders including leukemia, sickle cell anemia, or thalassemia, require regular red blood cell transfusions, and especially the care for the chronically transfusion-dependent generates substantial cost (Wayne2000). Likewise, patients with certain hematologic disorders including acute leukemias and certain cases of myelodysplatic syndrome who develop thrombocytopenia require extensive platelet transfusion support, once again at substantial expense (Meehan2000). Periodic transfusion often leads to progressive alloimmunization against an increasing number of antigenic determinants displayed on the donor cells, be they red cells (Castro2002) or platelets (TRAP1997).

Platelets: Human Leukocyte Antigens ("HLA") Class I, Human Platelet Antigens ("HP A") -

Patients receiving therapy for hematologic malignancies consume more than 40% of the approximately 2.1 million single donor units (or equivalents) collected in the US as of 2013 (AABB2013) . Many patients have antibodies, formed in response to prior allogeneic exposure during pregnancy or previous transfusion, and others develop antibodies during treatment, and these antibodies mediate the accelerated clearance of transfused cells, leading to a poor response to transfusion and excess platelet consumption as well as excess utilization of clinical services, and extended in-hospital stays, especially for patients who respond to transfusion (Meehan2000).

In part, this state of affairs reflects the logistical difficulty of identifying suitable platelet donors quickly in view of the short expiration dating for platelets. Random searches that identify prospective donors by a negative serological cross-match, are time-consuming and, at best, will exclude as unsuitable only those prospective donors with cognate epitopes to existing antibodies, but will not identify allo-epitopes that may lead to the formation of new antibodies. The genotyping of HLA, though long since a standard approach to matching stem cell recipients and donors, as currently practiced, is complex and slow , and the prevailing strategy of procuring stem cells has been to maintain large registries of volunteers who are genotyped at registration, an expensive propositions the vast majority of these volunteers will never called. Creating large registries of potential platelet donors clearly is impractical for the routine procurement of suitable platelets, given the large demand, and the time constraints imposed by the platelet expiration dating.

Red Blood Cells: Human Erythrocyte Antigens ("HEA ") - For sickle cell patients, stroke is a major risk factor, and timely (hence chronic) transfusion has been shown to be very effective in reducing that risk (eg. Lee2006). The commercial introduction of routine genotyping into donor centers and hospital transfusion services, a decade ago (Hashmi2005, Hashmi2007, Moulds2011), has greatly facilitated the procurement of suitable red cells especially for transfusion-dependent patients with multiple antibodies, a common side effect of chronic transfusion of sickle cell anemia and thalassemia patients (Castro2002, Pham2011,Chou2013) . However, notwithstanding its commercial availability in several formats, genotyping, given its perceived high cost and complexity - which may require special training and in some cases certification— has been limited, in practice, to special situations that are not readily handled by serology. Serology, largely automated, has otherwise remained the "work horse" in the pre-transfusion setting, especially for large-scale "pre- selection" of candidate donors.

Finding candidate cells with desirable molecular attributes, usually in the form of a set of cell surface markers (expressed or not expressed), or a set of antigenic determinants associated with antigens such as HLA, is a search problem. The prevalent format of genotyping represents a "brute force" solution that is ill-suited to scale up. Thus, to identify, in accordance with this format, donors who do not express the RBC antigens E, V and Fya, say, one first genotypes all candidate donors at hand - one at a time - for an entire set of alleles (as in, say, Bio Array Solutions' "HEA PreciseType" test, see website at Immucor, Inc.), then looks for instances, if any, that lack the specified antigens. As this attribute pattern - "E- & V- & Fya-" - is not a common one, many of the genotype determinations will be of no value, and unless they address other instances of pending requests, the investment made in those determinations may be lost. Consequently, to reduce genotyping expenditures, many hospital transfusion services supporting transfusion-dependent patients have resorted to extraordinary measures such as pairing individual patients with special ("buddy") donors on whose continuing kindness they count for a vital part of their patient care; all the while generating tens of thousands of dollars in annual expenses for other aspects of care for the very same patients, including laboratory charges or "spend" for iron chelators (Wayne2000).

In the hematology/oncology setting, the situation is worse. Unless patients actually become non- responsive to platelet transfusion, the procurement and selection of platelets, in order to ensure hemostasis and to maintain vascular integrity, remains largely uninformed by concerns about the risk of allogeneic exposure to antigens displayed on platelets, notably HLA (class I) and HPA, and its clinical and financial consequences

On the supply side, many practitioners rely on serological methods, preferably in an automated format, to "pre-screen" candidate units for genotyping, the latter usually extended to a small fraction generally not exceeding a few percent of the units collected. However, for all but the most common red cell antigens such as C, E and K, this approach must rely on a limited (and expensive) supply of reagents. In addition, it has the disadvantage that it not only proceeds one sample at a time, but also one antigen at a time, and therefore requires elaborate sample handling and tracking. Many practitioners also invoke simple heuristics for pre- selection, for example, on the basis of major blood type and/or declared ethnic background. Many also favor repeat donors, thereby in some cases severely narrowing the distribution of available antigen profiles.

To overcome the limitations of current approaches to large-scale genotyping generally, and to the routine procurement of blood cells or other cells with specific antigen and genotype profiles, a process is needed that: (i) enables the effective scale-up of genotyping to survey and profile large numbers of samples, and (ii) does so in a manner ensuring superior performance over the prevailing "brute-force" search strategy, where performance may reflect the number of operations performed, and time to completion of these operations, and related cost per "hit." The effective search process disclosed herein is related to "Nucleic Acid Sieving" (US Pat. No. 8,932,989 and US Publ'n No. 2015/0315568), "Allele Profiling" (US Publ'n Nos. 20130029857 and 2015/0376693) and "Attribute Profiling" (US Pat. No. 9,133,567), all of which are incorporated by reference, is disclosed herein. Summary

The new search process achieves its effectiveness and scalability by inspecting pooled samples for one or more sample attributes or combinations of sample attributes. Pools of samples are formed, as disclosed herein, in a manner reflecting expected abundances of desirable and/or designated sample attributes, wherein the attributes preferably are molecular attributes including: alleles or haplotypes; cell surface markers, including antigens, and especially antigens or epitopes recognized by antibodies, including those previously identified in intended recipients of red blood cells or other cells. "Desirable attributes" (note that, "designated attributes," can also be "desirable attributes," and "desirable attributes" can also be any attributes which are also attributes included in: "attributes of interest," "attribute configuration," "attribute set," or "attribute pattern") may or may not be directly detectable by the testing method employed, so the process also can be effected by monitoring/testing for certain detectable attributes whose presence or absence correlates with the presence or absence of desirable, designated or detectable attributes, as in the case of an attribute pattern comprising antigens whose presence or absence correlates with the presence or absence of the alleles encoding those antigens.

The preferred embodiment includes performing at least the first stage, sieving or panning stage, which includes the steps of: concurrent determination of multiple attributes or attribute patterns for pooled samples; the analysis of data recorded from the pools; and the selection of pools (and their constituent samples) by one or more selection criteria (aka "strategies" or "policies"), as illustrated in the Examples; the pool data analysis comprises identifying pools as unambiguous or ambiguous for desirable attributes or attribute patterns; one embodiment of a selection strategy is to enrich the selection for samples that have, or may have, desirable attributes or attribute patterns, including estimating the number of samples in ambiguous pools having one or more particular attributes, without resolution of ambiguities or molecular analysis of samples.

Optional additional stages include:

A profiling stage, for a subset of samples selected from unambiguous or ambiguous pools, comprising the resolution of any ambiguities remaining after the sieving/enrichment stage, and optionally, the determination of any additional attributes. Where, as in a preferred embodiment, both sieving and profiling steps are performed by DNA analysis, the profiling step, performed for selected samples, may confirm the partial profiles determined in the sieving step.

The process includes forming pools and optionally pooled pools, by combining aliquots from "d" samples, such that any sample of interest will be added to at least one pool and pools are unique; "d" is determined as a positive integer value that maximizes, or approximately maximizes, the expected number of samples in pools whose constituent samples are identical with respect to designated attributes, or a designated attribute set. There is an upper limit, d_max , on the total number d of samples per pool, where d_max is based on, among other things, the ability of the assay or testing technology to detect attributes at d_max - fold dilution.

The process may further include associating all samples in pooled pools with a particular pool, as well as identifying samples having particular attributes, preferably by forming, for each pool or pooled pool, attribute- specific, source-tagged reaction products wherein each such reaction product has a marker-tag identifying the attribute and a source tag identifying the pool , as described, e.g., in US Pat No. 8,932,989 and other references in the Background section, and as further described and defined below.

While many of the illustrative examples invoke the case of red blood cell ("RBC") antigens (aka "HEA"), the process and methods apply equally to selection by attribute patterns comprising other cell surface markers (e.g., platelet antigens HPA, or HLA), or other molecular attributes encoded by a set of known alleles (e.g. those for enzyme mutations or receptor polymorphisms), as well as to sample selection by allele patterns.

A pool is said to be unambiguous for a desired attribute pattern if its constituent samples are identical with respect to that attribute pattern where a pool may be simultaneously unambiguous for one or more of the attributes of interest, but ambiguous for others. For unambiguous pools, the analysis can stop after the first stage with the selection of the constituent samples for the designated or other desired attributes, as illustrated herein, or can be continued, for some or all of the constituent samples, in order to determine additional attributes that may not be shared by all the constituent samples. A pool is said to be ambiguous for a desired attribute pattern if its constituent samples are not identical with respect to that attribute pattern, but at least one constituent sample has, or may have, that attribute pattern; in one embodiment, the process comprises the analysis of pool data to estimate how many constituent samples are likely to have the pattern. Analysis generally should be continued to determine which (if any) sample or samples in the pool have the pattern.

The determination of attributes at stage 1 or stage 2 above can be made by methods of DNA analysis to determine alleles (by, for example, allele- specific amplification and sequencing, as in US Pat No. 8,932,989 and other references in the Background section or by other standard methods of nucleic acid analysis) and to predict expressed antigens, or directly by standard methods of immunopheno typing, to determine antigen type and/or by serology, to establish cross-reactivity between donor cells and prospective recipient serum.

Especially in the context of allocating suitable red blood cells to transfusion-dependent patients, reliance on serological methods and processes for the pre- selection step or otherwise has several disadvantages that are overcome by DNA analysis in accordance with the process disclosed herein. First, serological reagents for many red cell antigens (e.g. V, VS, Hy, Joa) are in short supply, or unavailable for routine use. Second, the use of serological reagents for identifying the important RH antigens C, c, E and e has been amply demonstrated to be unreliable when encountering RHCE variants, with serious adverse clinical consequences for transfusion-dependent patients (Pham2011, Chou2013). The principal advantage of the new process is its capacity for rapidly sampling a substantially larger fraction of available candidate donors, for a larger number of attributes, than that accessible by even fully automated serological "screening," which generally proceeds one sample at a time, one antigen at a time. The new process quickly and readily identifies samples with a number of desired attributes. The selection of samples, by matching partial profiles determined in the sieving step to desired attribute patterns, enriches - and the "de-"selection of samples not having desired attribute patterns depletes - the set of samples committed to any subsequent profiling step(s), as in step 2 above, thereby optimizing the benefit of additional analysis.. For example, if one starts with 384 samples, rare configurations are more likely to be detected if, for example, one has 96 pools of 4 samples each, as compared to randomly selecting sets of 96 samples. Thus, an advantage of the process disclosed herein is that of monitoring samples for multiple rare variants simultaneously, while concurrently permitting enrichment by selection for yet additional attribute patterns. In a preferred embodiment, the classification and pre- selection, by "sieving", serves to "tune" the composition of the sample sets selected for profiling in accordance with attribute patterns reflecting pending or anticipated demand.

Brief Description of the Drawings

Fig. 1 is a process flow chart showing the steps in the process of attribute profiling with pooling as described herein.

Fig. 2 is a flow chart showing how to determine "d" of Fig. 1.

Fig. 3 is a table illustrating the determination of "d", for the attribute pattern "E- & Fya-", using phenotype frequencies for African Americans (Reid2004); column 3 holds the values for the probability that all d samples are "C-", column 4 those that all samples are "Fya-". Notation: "Cau" denotes Caucasian; "Afa" denotes African Black; "NA" signifies a field to accommodate an additional marker; "Nof ' is short for "Number of.

Fig. 4 is a plot of the expected number of samples in pools unambiguous with respect to "E- & Fya-" (right-most column in the left-hand Table of Fig. 3) vs "d", the number of samples per pool (left-most column of the left-hand Table of Fig. 3). The plot displays a maximum at d=3.

Fig. 5 is a table reporting results of matching requests for sickle cell anemia patients with observed patterns of antibodies to red blood cell antigens (in accordance with Table 1 in Castro2002) with "antigen-negative" candidate donor cells selected from unambiguous pools. . For each unique observed antibody pattern, shown in the left-most column, the right-most column reports the expected number of samples found in 4-sample pools (that is: d=4) that were unambiguous with respect to that pattern, assuming phenotype frequencies for African American candidate donors (without consideration of "ABO/RhD" type); "mean" and "standard deviation" were computed for 10 replicates of 96 simulated 4-sample pools. Entries of "0", in the right-most column, indicate that no unambiguous pool was identified in the 10*96 pools.

Fig. 6 is a table illustrating selections of samples from 4-sample pools identified as unambiguous for attribute configurations "V- & VS-", "C- & V- & VS-" and "E- & V- & VS-"

Fig. 7 is a table illustrating selections of samples from three 4-sample pools identified, by mixed "all" and "any" queries, as ambiguous for attribute configurations "all C- & all V- and any hrB-". Profiling of the samples in these three pools would reveal the displayed phenotypes which indicate that: sample 1 in pool 1, sample 3 in pool 3 and sample 2 in pool 4, in addition to being "C- & V- ", as specified, also are "hrB-", and in fact the first and the last of these samples also are "E-."

Fig. 8 is a bar graph illustrating the presence of the variant allele, YT*02, in pool "p2"; pools "pi", "p3" and "p4" are negative for the variant allele, hence homozygous and thus "unambiguous" for the normal allele, YT*01. The increase in signal intensity recorded for pool "p2" over that of the others is indicated by the increased size of the bar.

Fig. 9 is a bar graph illustrating the detection of a variant form of the Single Nucleotide Polymorphism, 1006T (corresponding to 336C), associated with the alleles RHCE*01:20:03 and RHCE*01:20:05, neither of which express the antigen V. Pool "p3" is positive for the variant form of that SNP, indicating the presence of at least one copy of one of the variant alleles, while pools "pi", "p2" and "p4" are negative for the variant, indicating expression of the V antigen by all constituent samples. The difference in signal intensity of the p3 pool is shown by the increased size of the bar.

Fig. 10 is a bar graph illustrating the absence of certain HLA-B alleles comprising either "B 120A" or "B 120G" for four pools, each comprising four samples, with pairs of bars indicating the relative signal intensities for B 120A and B 120G for individual pools: no signal for "B 120A" is detected in pools 1 and 2.

Fig. 11 is an illustration of cis- and trans-configurations of a pair mutations, showing the positions of forward and reverse primers for detecting any cis-specific product.

Detailed Description

Definitions

"Desirable attribute" shall mean any attribute sought in a sample or samples.

"Designated attribute" shall mean any attribute designated in a sample or samples.

"Attributes of interest," "attribute configuration," "attribute set," and "attribute pattern" all refer to a set or subset of designated or desirable attributes. "Enriching" shall mean: increasing the abundance of samples in the set having desirable attributes over that achieved by random sampling of a population.

"Tag" refers to a molecule or portion thereof with a recognizable feature that allows it to be distinguished from other tags, or tag molecules, e.g., a distinguishable nucleotide or amino acid sequence, nucleotide or amino acid sequence length, shape, size, mass, color, optical density, differential absorbance or emission of light, chemical reactivity, magnetic or electronic properties and the like. Preferred examples of tags include tags comprising oligonucleotides (oligonucleotide tags) and fluorophores.. A specific oligonucleotide tag may serve to identify a sample or sequence, in the manner of a "barcode." A "tag" may include a florescent label so that the tag may be identified.

"Source tag" or "S-tag" is a tag that is attached to or comprises a polynucleotide or oligonucleotide and identifies the source of the polynucleotide or oligonucleotide or nucleic acid under study. In some embodiments, a source tag is an "oligonucleotide tag." Oligonucleotide tags may be identified by their nucleotide sequence, partial sequence (or composition) or length, or configuration or electrophoretic mobility.;. In some embodiments the oligonucleotide tag is a sequence of nucleotides selected such that the sequence does not duplicate a naturally occurring sequence in the genome of the organism under study; such an oligonucleotide tag also may also be referred to as a "barcode."

"Marker tag," or "attribute tag," as used herein, is a tag that uniquely identifies a polymorphic site and/or allele. In some embodiments, a marker or attribute tag is a tag that is attached to or comprises a polynucleotide or oligonucleotide and identifies an allele and/or polymorphic site under study. Oligonucleotide tags may be referred to as "barcodes." In some embodiments, the marker or attribute tag may identify an allele and/or polymorphic site under study by the length or sequence of the oligonucleotide tag. In some embodiments, the marker or attribute tag may identify an allele and/or polymorphic site by a fluorescent label.

Referring to Fig. 1, the first process steps are: selecting desirable and designated desirable attributes, and based on the known or estimated frequencies of the designated attributes in the population; and determining the preferred or optimized number, "d", of samples per pool. Figs. 2 to 4 more particularly illustrate applying the algorithm for determining "d". Next, sample pools are formed from a total of N samples, generally with "d" samples in each pool, resulting in N/d pools, where N preferably is chosen to produce an integer for N/d. Samples, or sample-derived products in the pools are then source-tagged ("S-tagged"), preferably in the process of forming first reaction products, to indicate their pool of origin, for identification of that pool of origin in pooled pools that may be formed in further steps, as shown in the far right of Fig. 1, prior to forming attribute- specific source-tagged second reaction products. The attributes of interest displayed by S-tagged products are attribute-tagged, where the attribute tagging preferably uses a fluorescent or other color tag or sequence tag, and where the attribute-tag identifies antigens or alleles. Samples placed into sample pools may be obtained from sources including: whole blood (comprising circulating blood cells including stem cells); lysates produced from buccal swabs (comprising epithelial cells) or saliva; or purified genomic DNA extracted from nucleated cells, or freely circulating DNA or other nucleic acids. The next step is to establish selection criteria for the S-tagged attribute-specific reaction products in the pools or pooled pools (see Examples below), where the selection criteria can refer to, e.g., all of a particular set of attributes of interest, or combinations of any such attributes. Pools with the desirable attributes are selected (or deselected),; other pools may be discarded. Among selected pools may be pools that are unambiguous for the desirable attributes, so that, even at this stage, the attribute profile of all constituent samples is known at all the positions corresponding to these attributes. Optionally, the attribute profiles of samples in unambiguous pools can be left incomplete (assigning "wildcards" to positions left undetermined, as in Fig. 1), or the samples may be individually profiled to determine any additional attributes of interest. Among selected pools also may be pools that are ambiguous for the desirable attribute(s), that is, they contain attribute- specific reaction products having the desired attribute(s), as well as others that do not; for example, pools which show both the color tag(s) associated with a normal as well as the color tag(s) associated with a corresponding variant allele. In such a situation, ambiguity arises as it is unknown which of the samples in such pools have, or may have, the desired attribute(s), and which do not. To eliminate (or reduce) ambiguities, samples in ambiguous pools can be separately assayed (or first assayed in groups, to save time) for the attribute- specific properties of interest. In one embodiment of selecting pools, in order to guide the selection of particular ambiguous pools over others, an estimate is made of the likelihood of how many samples in the pools include the desirable attribute and pool data are analyzed to identify corresponding subsets of ambiguous pools. Fig. 2 illustrates determining "d", given desired attributes and their frequency in the population of interest. In this case, the desirable attribute(s) can mean the presence or absence of an antigen, or antigen combinations, or the presence of normal ("N") or variant ("V") alleles. In Fig. 2, the desirable attribute is: "negative for red blood cell antigens E and Fya", denoted by "E- & Fya-" (short-hand for "all E- & all Fya-").

To determine the preferred "d", determine the probability, as a function of increasing positive integer values of "d", that all samples in a pool of d samples (aka d-sample pool) have the designated attributes interest (which may mean having the attributes , or lacking them, so that, in either case, the pool is unambiguous for the attributes); in this case, this is the probability that all samples lack the antigens E and Fya. Next, multiply that probability by the anticipated number of pools to be formed, to obtain the expected number of pools that are unambiguous for the designated attributes (that js: the number of pools with "no variant attribute"), and multiply that number by "d" to obtain the corresponding expected number of samples in such unambiguous pools. This number will display a maximum, at some value of "d", because, as "d" increases, the number of samples per pool increases, but the expected number of pools remaining free of at least one variant decreases. This is shown in the table in Fig. 3: for increasing values of "d", in the left-most column, the right-most column recording the expected number of samples in d-sample pools that are unambiguous with respect to the pattern "E- & Fya". Fig 4 is a related plot which displays a maximum at d=3. Thus, the position of the maximum, and the optimal value of "d", may be determined from this table or the corresponding plot. To summarize, with reference to the table in Fig. 3, determine "d" as follows. Given an attribute pattern, and the frequencies of the constituent attributes, compute, the probability that all "d" samples have that attribute (where having the attribute can mean expressing certain antigens, or NOT expressing certain antigens, as in the table in Fig. 3). Thus, for the attribute pattern "E- & Fya-", that is: "E negative and Fya negative", with respective abundances FE = 22% (that is: 22% of individuals express the antigen E and thus have the phenotype E positive) and FF_ya = 10%, as reported for African Americans (Reid2004), the probability that a sample will be E negative AND Fya negative is (l-FE)*(l-FF_ya), and the probability that all samples in a pool of d-samples will have that attribute pattern is prob(d) = (l-FE)^D * (l-FFya)^d . When referring to the frequencies of the encoding alleles (here comprising the 2-valued single nucleotide polymorphisms RH676G>C and FY125A>G), rather than to phenotype frequencies, the corresponding probability of not having any variant allele , is (1- fRH676c)²*^d * (l-fFY125A)²*^d .

Thus, to obtain the expected number of pools, among n such pools, multiply that probability by n, say 96; to obtain the expected number of samples in unambiguous pools, multiply that probability by d*96: this number displays a maximum at some value of "d", here at d=3 (Figs 3, 4). Set the desired "d" to a positive integer value at or near the position of the maximum, for example d=4: at the cost of a minor decrease in the maximum value of expected samples in unambiguous pools (i.e., 96 at d=3 vs 92 at d=4,) selecting the larger d-value permits surveying a larger number of initial samples, d*96. Further, combining samples in accordance with powers of 2 has certain practical advantages in the laboratory as it facilitates adoption of recursive processes for making pools, e.g. "pair samples", then "pair pairs of samples" and so on.

The process accommodates two or more designated attribute patterns of interest, as follows. For each such attribute pattern, determine the preferred "d" as described. If these first and second preferred values of "d", say "dl" and "d2" differ, and especially if they differ substantially, dl » d2, say, prepare two sets of pools accordingly, one with pools comprising "dl" samples each, the other with pools comprising "d2" samples each, and perform the "sieving" step for each pool set, in parallel or in series. In the former case, the "d2"-pools may hold samples randomly selected from those placed into the "dl"-pools or may comprise a separate set of samples; in the latter case, pools placed into the "d2"-pools may be enriched by samples selected from selected "dl"-pools.

Examples

Simulation of Antigen Profiles in Random 4-Sample Pools - A set of 96 pools, each comprising d=4 samples was generated, each of the samples represented by a phenotype constructed by randomly selecting pairs of alleles encoding the principal antigens associated with the RBC blood groups MNS, RH, LU ("Lutheran"), KEL ("Kell"), FY ("Duffy"), JK ("Kidd"), Dl ("Diego"), YT ("Cartwright"), DO ("Dombrock") and CO ("Colton") (Reid 2004). Allele frequencies were determined by analysis of experimental genotype data for African Americans in the Southern United States. For illustration, a pool comprising a set of four phenotypes is as follows:

- [1] "M-s-U-c-e-hrB-hrS-Lub-Jsb-k-Kpb-Fya-Jka-Jkb-Dib-Yta-Dob-Hy-Joa-Coa"

- [2] "M-N-s-S-U-c-CW-e-hrB-hrS-Lub-Jsb-k-Kpb-Fya-Jka-Jkb-Dib-Yta-Doa-Hy-Joa-Coa" - [3] "M-s-S-U-c-e-hrB-hrS-Lub-Jsb-k-Kpb-Fya-Jka-Jkb-Dib-Yta-Dob-Hy-Joa-Coa"

- [4] "M-s-U-c-e-hrB-hrS-Lub-Jsb-k-Kpb-FyNull-Jkb-Dib-Yta-Doa-Dob-Hy-Joa-Coa"

The information from this simulation was used in generating the information in several of the examples below.

"All" Queries (see also below ) - These queries identify pools where all samples are identical with respect to the specified attribute pattern, e.g.: identify pools wherein ALL samples lack the antigens C, E and K: "C- & E- & K-" (short for "all C- & all E- & all K-"); or "C- & E- & CW-"; or "E- & Fya- & Jkb-", etc.; pools identified by such "all" queries are said to be "unambiguous" with respect to the specified attribute pattern.

Filling Requests by Selecting Donor Units from Unambiguous Pools: Sieving Only - The table in Fig. 5 reports allo-antibody combinations observed in transfusion-dependent sickle cell anemia patients (extracted from Table 1 in Castro2002). Transfusion support for these patients calls for the procurement of "antigen-negative" units, that is: units that do not expose patients with allo- antibodies to cognate antigens. "All" queries, against 10 sets of 96 four-sample pools generated by simulation, return, for attribute patterns comprising the set of cognate antigens for each of the observed antibody combinations in the second column off the table in Fig. 5, the number of pools determined to be unambiguous for that combination: the mean over 10 replicates, and the corresponding standard deviation are shown along with the expected number of samples in unambiguous pools, in the right-most column. Thus, unambiguous pools would yield, without the need for further analysis, suitable samples for 103 of the 135 patients included in the compilation (though total consumption, not provided in Castro2002, may be expected to vary from patient-to- patient); this specifically includes most patients with up to 3 antibodies, and many of those with up to 5 antibodies. In addition, the set of samples selected for further analysis may be enriched by samples with specific partial attribute profiles. In many cases, pools selected as unambiguous for a designated attribute pattern, also are unambiguous with respect to additional attributes of potential interest. For example, in 10 sets of 96 four-sample pools, those selected as unambiguous with respect to "C- & K- & Jsa- & Fyb- & S-", on average 4.4 (+/- 1.79) pools, also were unambiguous for "Dia- & Cob- & Ytb- & Joa+ & hrS+", and all but 3 (of 44) pools, in addition, were unambiguous for "CW- & CX-", while 18 were unambiguous for "M+", so that, for a small number of pools, even the MNS antigens were determined without the further analysis. Thus, sieving alone furnishes a considerable amount of information, even for complex desired attribute patterns.

Sample Allocation Policy - The pools identified in the table in Fig. 5 as unambiguous for the absence of antigens targeted by antibodies commonly observed in sickle cell patients are overlapping pools. That is, the same pool may be unambiguous for more than one of these attribute patterns of interest, and this calls for an allocation strategy to determine which of two or more desired attribute patterns will govern the selection of samples from specific qualifying pools. Generally, the more challenging an attribute pattern, the smaller the expected number of suitable candidate samples, suggesting a first allocation strategy in accordance with this greedy heuristic: select samples for the most challenging attribute pattern first. Here, challenging attribute patterns are those producing a high probability of an ambiguity (thereby reducing the expected number of pools remaining free of ambiguities); thus, the larger the number attributes in a pattern, or the more polymorphic an allele encoding antithetical antigens, say, the more challenging the pattern comprising those attributes. For example, with reference to the table in Fig, 5, having identified, in a simulated set of 96 four-sample pools comprising red cell phenotypes, pools 7,42,44,48,89 and 95 as unambiguous for "C- & E- & Fya-", pools 48 and 89 as unambiguous for "C- & E- & K- & Fya- & S-" and pools 7,40 and 95 as unambiguous for "C- & Fya- & Fyb-", then, from the larger pool set, samples in pools 7 and 95 would be allocated to recipients having antibodies to "C, Fya, Fyb"; samples in pools 48 and 89 to recipients having antibodies to "C, E, K, Fya, S" and samples in the remaining pools 42 and 44 to recipients having antibodies (only) to "C, E, Fya". Many alternative heuristics may be invoked to guide the allocation of samples or selection of qualifying pools.

Ignoring Certain Antithetical Antigens - The selection of candidate donor samples may be guided by the antigen profile of intended recipients; for example, selection may be in accordance with combinations of known antibodies, as in the table in Fig. 5 In the case of certain antithetical antigens, if a recipient expressed both antithetical antigens, reflecting heterozygosity for the encoding alleles, then, no harm is done regardless of the values of any candidate donor sample at these antithetical antigen positions within a molecular attribute profile, that is: whether the donor sample is heterozygous or homozygous makes no difference, and such attributes can safely be ignored in selecting suitable donor samples. "Any" ("At Least One") Queries - These queries identify pools comprising at least one sample having (or not having) the specified attribute pattern; for example, to identify all pools wherein at least one sample lacks the antigen "c": "any c-"; or "any e-", or "any hrB- & hrS-" or "any Lua+"; "any" queries may be combined with "all" queries, e.g.: "all C- & all E- & all K-" & any s-"; or " all C- & all K- & all Fya- & any e-" & any S-". Pools identified by "any" queries are said to be ambiguous for the specified attribute(s); an example is as follows.

Identifying Uncommon Antigen-Negative Configurations: Ambiguous Pools - Patients with certain RHCE variant allele combinations will not express the antigen hrB, say, and may develop allo-antibodies as a result of exposure to that antigen, commonly expressed by Caucasian red cell donors. Continued transfusion support then calls for hrB- ("hrB neg") donor units. Pools comprising at least one sample lacking hrB may be identified by an "any" query, which may return a pool comprising, for example, these phenotypes:

[1] "N-s-U-c-e-hrB-hrS-VS-Lub-Jsb-k-Kpb-FyNull-Jka-Jkb-Dib-Yta-Dob-Hy-Joa-Coa"

[2] "M-N-s-U-c-e-hrB-hrS-VS-Lub-Jsb-k-Kpb-Fya-Jka-Jkb-Dib-Yta-Doa-Hy-Joa-Coa"

[3] "N-s-S-U-C-e-Lub-Jsb-k-Kpb-FyNull-Jka-Dib-Yta-Dob-Hy-Joa-Coa"

[4] "M-N-s-U-c-cE-e-E-hrS-V-VS-Lub-Jsb-k-Kpb-FyNull-Jka-Dib-Yta-Doa-Dob-Hy-Jo a-Coa"

Samples 3 and 4 lack hrB. However, to identify these samples within this "ambiguous" pool requires disambiguation, preferably by "profiling" of individual samples (d=l): by producing the complete attribute profile for each constituent sample, profiling also may "fill in" any gaps left in the partial profiles identified by sieving for a specific attribute pattern.

When constructing queries to identify pools that are ambiguous for more than a single attribute, it must be borne in mind that the specified attributes are associated with pools, NOT (necessarily) with each constituent sample. Thus, a query for pools with at least one sample having "hrB- & hrS-" may return a pool such as the one above comprising: samples 1 and 2, both expressing hrB and hrS, sample 3, expressing neither hrB nor hrS, and sample 4 expressing hrS but not hrB.

Confirmation of Homozygosity by Pooled Serology - Given a pool that is ambiguous with respect to the alleles encoding antithetical antigens, such as "e" and "E" (above), with at least one variant allele, serology may be used to confirm the presence of any homozygous sample. To that end, perform an immunoassay with a pool of cells from the sources of the DNA analyzed, and expose the cells to anti-e antibodies, labeled with a first fluorescent dye ("RED", say) and to anti-E antibodies, labeled with a second fluorescent dye ("GREEN", say). Cells from a heterozygous sample will bind both RED and GREEN antibodies, cells from homozygous samples will bind antibodies of only one color If indeed present, a true "e-" cell type will bind only the GREEN (anti-E) antibody. This immunoassay format thus conveniently distinguishes between ambiguous pool configurations reflecting the presence of heterozygotes and those reflecting the presence of at least one true homozygous variant, often the object of interest. Flow-cytometric analysis with at least two color channels, would provide convenient implementation, by gating on events comprising a single dye or both dyes. To limit profiling to pools containing homozygous variants, pools not containing true homozygous variants may be excluded from further analysis.

In one embodiment, data recoded from ambiguous pools are used to identify distinct subpopulations of ambiguous pools that differ in the respective number of constituent e and E alleles (or other antithetical antigens) and thereby narrow the selection of ambiguous pools to those most likely containing one or more constituent samples homozygous for "e-" (and by analogy homozygous negative for other common antigens).

"Sieving" Samples for Desirable Attribute Patterns: Enrichment - African Americans display significant allelic diversity especially with respect to RHCE variants and corresponding antigen profiles. Notably, the antigens V and VS, encoded by RHCE variants RHCE*01:20, at respective abundances of 30% and 26-40%, are sufficiently common in African Americans that, when selecting a suitable African American donor for a patient lacking, say C, V and VS, or E V and VS, the candidate donor must be confirmed not to express C, V and VS or E, V and VS. While it may be tempting, in this situation, to simply rely on donors of Caucasian ethnic background who only rarely express these antigens, genetic differences with respect to other blood group antigens, notably Duffy, foreclose that option. To wit: approximately, 70% of African Americans are homozygous for the allele FY*02N:01 which silences the expression of the antigen Fyb. These individuals display the phenotype FyNull and therefore would be exposed, with near certainty, to Duffy antigens, Fyb or Fya, if given a unit from a donor of Caucasian or other non-African ethnic background. The method disclosed herein provides an effective means of accomplishing the objective of identifying suitable African American donors, by enriching the set of candidate donors selected for profiling of red blood cell antigens or their coding alleles.

Simulations of ten replicates of 96 four-sample pools, using population frequencies determined from experimental genotyping data for African Americans in the Southern United States, show the respective mean population frequencies for "V+" and "VS+" in randomly selected sets of 96 African American phenotypes, to be -33% and 40%.

Desirable donors would be (at a minimum) "V-", or "VS-"or preferably "V- & VS-". Fig 6summarizes outputs from simulated sieving runs, each comprising 10 replicates of 96 four- sample pools, and shows: an average of: -38.8 samples in -10 pools identified as unambiguous for the attribute pattern "V- & VS-" ; -8.8 samples in -2 pools identified as unambiguous for "C- & V- & VS-" ; -10.4 samples, in -2.5 pools identified as unambiguous for "E- & V- & VS-"; and -15 samples in pools identified as unambiguous for "C- & V- & VS-" or "E- & V- & VS-".

Substituting samples in these unambiguous pools for randomly selected samples, in the set to be committed to profiling will substantially enrich the profiled set for the selected attribute patterns, as illustrated in this table:

Thus, a single sieving pass, comprising 96 four-sample pools, will enrich a profiled set of 96 samples by an expected 17.2% for samples having the attribute pattern "C- & V- & VS-" ; a second sieving pass will further increase the enrichment, as shown. An additional advantage of the method disclosed herein is this: enrichments for multiple attribute patterns may be achieved simultaneously, drawing on pools identified as unambiguous for desired attribute patterns, as well as on ambiguous pools that may contain variant alleles; the latter illustrated in the following Examples.

Detecting Variant Alleles in Pooled Pools - The routine detection of variant alleles, and especially rare variant alleles, in the sieving stage, as enabled by the method disclosed here, permits the sampling of a large number of candidate samples for several rare antigen configurations, even and especially when serology reagents are not available, as in the case of many less common antigens, e.g. those in the DO or YT groups or less common antigens in the RH group, as illustrated here with reference to Figs. 7, 8 and 9.

In rare cases, patients lacking a common (aka "high incidence") antigen such as "k" or "Yta" or "Lub", when exposed to the antigen during transfusion - almost a certainty given the high incidence of the antigen they lack - may form an antibody. Such antibodies directed against a common antigen substantially complicate any subsequent transfusion support which now calls for donor units that likewise lack a common antigen. To identify such rare donor units, it is helpful to "cast a wide net" by surveying a large number of candidate units for "het" configurations, indicated by at least one variant allele such as KEL*01, LU*01, or YT*02, as illustrated in Fig. 8.

For antithetical antigens, an ambiguity implies the presence of at least one copy of the antithetical antigen. Thus, unless a pool is ambiguous for the attribute "E-", it cannot contain a sample that is homozygous for the attribute "e-", and thus an ambiguity is a necessary (but not sufficient) condition for the existence of such a sample. This will be of special interest when looking for samples that are homozygous for rare variants, such as RH-C^X, RH-C^W, Cob, Lua, Dia, DO-Joa, DO-Hy: pools that are NOT ambiguous for any of these attributes may be "de-selected" after the sieving stage. For such a search, the preferred "d" may be determined as described above, but using the probability of encountering at least one variant attribute; for example, a pool of d samples cannot contain an "e-" sample unless it contains at least one copy of "E", an event that has probability l-(l-FE)^d. When the abundance is small, that probability, and the corresponding preferred "d" will be large and in practice likely limited by d_max - for example, with d_max = 32, and FDia = 0.001, setting d to d_max would produce an expected number of 96*(l-(l-FDia)³²) = 3 ambiguous pools: only those would be candidates for further analysis. In other cases, especially for the predominantly African American sickle cell anemia population, patients may lack one or both of the antigens V and VS and may also lack one or both of the antigens hrB and hrS, encoded by certain variant RHCE alleles. These antigens are expressed not only by essentially all Caucasians, but also by a substantial fraction of African Americans (see also above), so that even, or especially, when candidate units are given by African American donors, the risk of allogeneic exposure to V, VS or to hrB and hrS antigens, is significant. Accordingly, to avoid such exposure, African American donors expressing these antigens should be identified. Thus - with reference to Fig. 7 - when looking for, say, candidate donors with the attribute pattern "C- & V- & hrB-", only phenotype 1 in pool "pi" and phenotype 4 in pool 2 may be suitable, though that determination requires profiling (d=l), as all pools in Fig. 7 are ambiguous for hrB-, as indicated by inspection of the allele G336C; see also Fig. 9.

Including or Excluding HLA Allele Groups - By querying pools for one or more specific single nucleotide polymorphism(s) (aka "SNP" or "SNPs") defining HLA alleles, especially if, based on population frequencies of the nucleotide observed at those specific SNP positions, these are expected to produce unambiguous pools, entire allele groups that share (or do not share) a particular nucleotide for the specific one or more SNPs may be selected ("included") or de-selected ("excluded").

For example, the SNP at position 120 in the B-locus bi-partitions the set of B-alleles into those having "B 120G" and those having "B 120A". For individuals of African descent, 80% of the alleles comprise the former, and 20% of the alleles the latter. The probability of encountering at least one of the less common alleles, in a pool containing d=4 samples, hence < 2*4 B-alleles, is: prob(no "B 120A" in pool) = (l-fB-i20A)²*^d = 0.167, yielding, an expected number of 16 (=96*prob(no "B 120A" in pool)) pools not containing the variant. Accordingly, for all samples in such pools, alleles comprising "B 120A", and thus all alleles in the groups B* 13, B*40, B*44, B*47 are eliminated. See IPD IMGT/HLA website.

Fig. 10 illustrates the experimental signature of this situation: pools 1 and 2 do not produce a signal for "B 120A" and thus are unambiguous for "B 120G", while pools 3 and 4, produce signals for both "B 120A" and "120G" and thus are ambiguous. Similar examples may be readily constructed for other variable positions, and for other loci. By "any" or "all" queries, pools may be selected for inclusion in, or exclusion from profiling. For example, given a requirement for a platelet donor with a specific HLA class I allele profile, comprising an allele in the group B*40, sieving will facilitate excluding unsuitable candidates from a large initial candidate set.

Building a demand-informed population of callable prospective donors of cells - Blood centers, such as those operated by the American Red Cross and hospital transfusion services, such as those operated by many US tertiary care medical centers, as well as stem cell registries, such as the National Marrow Donor Program, in order to establish populations of callable donors of red cells, platelets and stem cells, rely on randomly recruited volunteers. Especially in connection with red cell and platelet collection, repeat donation is highly valued as they lighten the burden of donor recruitment and, to the extent that historic records of donor molecular attribute profiles are available, reduces the expense of genotyping. However, as a method for establishing a callable inventory with a desired composition of molecular attribute profiles, random sampling represents an inefficient strategy even if anticipated recipients and prospective donors were drawn from the same population - generally not a valid assumption.

The method disclosed herein enables a "directed" sampling strategy of superior efficiency in building a demand-informed callable inventory. As illustrated in the foregoing Examples, e.g. with reference to Fig. 5, the new process may be used to enrich the candidate pool for desirable attribute patterns, by selecting samples for attribute patterns that reflect historical and/or anticipated demand. In comparison to the prevailing, de facto random sampling approach to recruiting cell donors, directed sampling will be particularly efficient in identifying less common attribute patterns.

Patient Enrollment for Clinical Trials: Stratification by Molecular Attribute Profile - A key aspect of personalized (aka precision) medicine is the association between the response of individual patients to therapy and the patients' relevant genetic attributes. Of particular interest in this context are alleles comprising receptor and enzyme polymorphisms. An example of the former is the set of polymorphisms of the ADRB2 gene encoding the p2-adrenergic receptor (Hizawa2011); an example of the latter is the set of polymorphisms affecting the catalytic activity of the enzymes in the cytochrome P450 superfamily, including CYP2D6, CYP2C9 and CYP2C19 (for clinical significance and other detail, see e.g. ARUP Labs - website), with attendant effects on individual rates of metabolizing prescription drugs {See Indiana University, Dept. of Medicine, P450 drug interaction table). The determination of individual genotypes and constituent alleles therefore may inform treatment decisions, including drug and dose selection, and may also guide the design of clinical trials (Kurose2012) including the stratification of cohorts by genotype.

For example, the pooling method disclosed herein may be used to stratify patients by CYP genotype(s) of interest. Thus, to classify, by one or more CYP genotypes, Caucasian candidate participants in a drug trial for, say, angiotensin II receptor antagonists, where the CYP2C9 genotypes comprise one or both of the two principal variant alleles CYP2C9*2 and CYP2C9*3, both of which confer a reduced metabolic rate, proceed in accordance with the method disclosed herein, as follows. Select as a desired attribute pattern the absence of the alleles CYP2C9*2 (identified by the SNP 430C>T, R114C, with a frequency of -14% in Caucasians); determine an optimal value yielding 228 alleles (or 114 samples, given bi-allelic CYP genotypes) in pools unambiguous for the absence of the designated alleles, and a near-optimal, preferred value of d*=8, yielding 226 alleles (or 113 samples) in pools unambiguous for the absence of the designated alleles. Likewise, with a desired attribute pattern defined by the absence of CYP2C9*2 as well as CYP2C9*3, (SNP 1075A>C, I359L, ~ 6.4%), determine an optimal value d_opt=4, yielding 160 alleles (or 80 samples) in pools unambiguous for the desired attribute pattern, and a near-optimal, preferred value of d*=8, yielding 128 alleles (or 64 samples) in pools unambiguous for the desired attribute pattern.

Mutation Carrier Screening: Phasing - In like manner, the method of the present invention can be used to expand and accelerate programs of mutation carrier screening, targeting, for example, mutations associated with elevated cancer risk (e.g. BRCA-1 and BRCA-2) or autosomal recessive disorders including hemochromatosis, cystic fibrosis, β-thalassemia, sickle cell disease, lysosomal storage diseases and others. Attribute patterns of interest will comprise one or more mutant alleles: a sample with at least one mutant allele will introduce ambiguity. As the population frequencies of mutant alleles generally are in the range of at most a few percent, the method disclosed herein will call for the testing of large-d pools and thus enable large-scale screening for entire sets of mutant alleles. Of particular interest especially to recessive autosomal disorders with many causative mutations including those mentioned above is the identification not only of individuals who carry one copy, or two copies, of a mutant allele, but also patients who are heterozygous for two or more mutations, a configuration also known as "compound heterozygotic", and specifically the identification of "cis" vs "trans" configurations: in the former, a single chromosome carries two mutations ("cis") and the other remains intact, while in the latter each chromosome carries one mutation ("trans"), leaving no intact allele - is has been well recognized that this distinction has significant implications for the type and severity of clinical symptoms. The method of the invention, because it produces allele- specific products, permits the identification of cis- vs trans-configurations of mutations. Specifically, a nested PCR design comprising allele- specific primers to both mutations, and one reverse allele- specific primer to the downstream mutation generates a product, of characteristic length D, only for the cis-configuration (as depicted in Fig. 11). The presence of at least one such cis- specific product in a pool may be detected by a preferred embodiment of the method of the present invention using capillary electrophoretic analysis of attribute specific reaction products, as is well known and is disclosed in US Publ'n No. 20150376693, incorporated by reference: the cis-specific product produces a peak at a specific design location, and so long as the distance, D, between the mutations exceeds a minimal length of, say, 50 nucleotides, this peak may be detected, in a pre-determined color channel, in pools or pooled pools of attribute specific reaction products.

The disclosures of each and every patent, patent application, and publication cited herein are hereby incorporated herein by reference in their entirety. While this invention has been disclosed with reference to specific embodiments, it is apparent that other embodiments and variations of this invention may be devised by others skilled in the ait without departing from the true spirit and scope of the invention. The appended claims and their equivalents are intended to be construed to include all such embodiments and equivalent variations.

Although described in connection with preferred embodiments thereof, it will be appreciated by those skilled in the art that additions, deletions, modifications, and substitutions not specifically described may be made without departure from the spirit and scope of the invention, which is only defined in the appended claims and not elsewhere. References

AABB2013 - "Patient Blood Management Survey Report", A ABB Dec 18, 2015.

Castro2002 - Castro X Sandler SG, Houston- Yu P, Rana S "Predicting the effect of transfusing only phenotype-matched RBCs to patients with sickle cell disease: theoretical and practical implications", Transfusion 2002 Jun;42(6):684-90.

Chou2013 - Chou S T, Jackson T, Vege S, Smith- Whitley K, Friedman D F & Westhoff C M "High prevalence of red cell alloimmunization in sickle cell disesase despite transfusion from Rh-matched minority donors", Blood 2013 Aug; 122(6): 1062-1071.

Doehner2010 - Doehner H, Estey E H, Amadori S, Appelbaum F R, Buechner T, Burnett A K et al "Diagnosis and management of acute myeloid leukemia in adults", Blood 2010; 115:453 474.

Hashmi2005 - Hashmi G, Shariff T, Seul M, Vissavajjhala P, Hue-Roye K, Charles-Pierre D, et al. "A flexible array format for large-scale, rapid blood group DNA typing", Transfusion. 2005 May; 45(5):680-8.

Hashmi2007 - Hashmi G¹, Shariff T, Zhang Y, Cristobal J, Chau C, Seul M, et al.,

"Determination of 24 minor red blood cell antigens for more than 2000 blood donors by high- throughput DNA analysis", Transfusion. 2007 Apr; 47(4):736-47.

Hizawa2011 - Hizawa N, "Pharmacogenetics of β2 agonists", Allergology International. 2011; 60:239-246

Kurose2012 - Kurose K, Sugiyama E & Saito Y, "Population Differences in Major Polymorphisms of Pharmacokinetics/Pharmacodynamics -related Genes in Eastern Asians and Europeans", Drug Metab. Pharmakokinet. 2012; 27(1): 9-54. Lee2006 - Lee M T, Piomelli S, Granger SS, Miller S T, Harkness S, Brambilla D J et al.

"Stroke Prevention Trial in Sickle Cell Anemia (STOP): extended follow-up and final results", Blood. 2006 Aug 1; 108(3): 847-852.

Meehan2000 - Meehan K R, Matias C O, Rathore S S, Sandler S G, Kallich J, LaBrecque J et al. "Platelet Transfusions: Utilization and Associated Costs in a Tertiary Care Hospital", Am J Hematology 2000; 64: 251-256.

Moulds2011 - "BeadChip Molecular Immunohematology", Moulds J M, Ness P M & Sloan S R, Eds, Springer Verlag, New York, 2011.

Pham2011 - Pham BN, Peyrard T, Juszczak G, Beolet M, Deram G, Martin-Blanc S, Dubeaux L Roussel M, Kappler-Gratias S, Gien D, Poupel S, Rouger P, Le Pennec PY, "Analysis of RhCE variants among 806 individuals in France: considerations for transfusion safety, with emphasis on patients with sickle cell disease", Transfusion. 2011 Jun; 51(6): 1249-60.

Reid2004 - - Reid M & Lomas Francis C, "The Blood Group Antigen Factsbook", 2^nd Ed, Elsevier 2004

M. Sobel, P.Groll, Group Testing To Eliminate Efficiently All Defectives in a Binomial Sample, Bell Labs Technical Journal 1959 Vol. 38, Issue 5, pp. 1179-1252

TRAP1997 - TRAP Study Group. Leukocyte reduction and ultraviolet B irradiation of platelets to prevent alloimmunization and refractoriness to platelet transfusions. The Trial to Reduce Alloimmunization to Platelets Study Group. New Engl J Med 1997; 337(26): 1861-69.

Wayne2000 - Wayne A S, Schoenike S E, Pegelow C H, "Financial analysis of chronic transfusion for stroke prevention in sickle cell disease", Blood 2000; 96(7): 2369 - 2372 (2000).

Claims

WHAT IS CLAIMED IS:

1. A method of selecting and enriching biological samples for desirable attributes by forming sample pools; determining some of said desirable attributes for pooled samples; and selecting samples appearing to have one or more desirable attributes, the process comprising: determining, for at least one of the desirable attributes, the preferred number, "d", of samples per pool by setting d to the integer value 2) which produces the maximum or a value near the maximum of the product of: d times the number of sample pools expected to be unambiguous for the at least one of the attributes; dividing the plurality of samples into subsets, each comprising approximately d samples per subset such that each of the plurality of samples is included in at least one subset, and, forming unique pools with aliquots from different sample subsets; forming, in the pools, attribute- specific reaction products for the constituent samples, each product having a marker-tag identifying a desired attribute and optionally a source-tag identifying the pool; identifying pools that are unambiguous for one or more of a first set of the desirable attributes or ambiguous for one or more of a second set of the desirable attributes wherein all samples in said unambiguous pools have an attribute in said first set or no samples do; and wherein at least one of the samples in each of said ambiguous pools has an attribute of said second set, but not all samples do; and identifying and selecting samples from the unambiguous or ambiguous pools to form a sample set that is enriched for said desirable attributes.

2. The method of Claim 1, further including, repeating the steps with additional desirable attributes in the first or second sets.

3. The process of Claim 1 wherein, in a first reaction, source-tagged reaction products are formed which are capable of being interrogated for the purpose of identifying the desirable attributes, and wherein, in a second reaction, attribute- specific source-tagged reaction products are formed, each such attribute- specific source-tagged reaction products comprising a combination of source- tag and marker- tag.

4. The process of Claim 1 further including analyzing samples from ambiguous pools to determine their attribute profiles for the first set, the second set or for all desirable attributes.

5. The process of Claim 1 wherein the constituent samples from ambiguous pools are further analyzed individually to resolve which samples have, and which do not have, the desirable attributes.

6. The process of Claim 1 wherein selected samples with desirable attributes also have attributes left undetermined, and wherein these attributes are ignored when selecting samples.

7. The process of Claim 6 wherein the desirable attributes are antithetical antigens and wherein for constituent samples of a pool that is ambiguous with respect to antithetical antigens, the presence of at least one sample that is homozygous for a desired antigen is confirmed by an assay comprising cells expressing the antigens, wherein a first antibody labeled with a first color is directed to a first of the antithetical antigens, and a second antibody labeled with a second color is directed to a second of the antithetical antigens, to determine the presence or absence, in the pool, of cells expressing both antithetical antigens.

8. The process of Claim 7 wherein the ambiguous pools include samples with both antithetical antigens which are not included in any selected samples.

9. The process of Claim 1 wherein a sample can be homozygous or heterozygous for the alleles encoding a desirable attribute

10. The process of Claim 1 wherein the value selected as "d" is greater than that yielding the maximum of the product of: d times the number of expected unambiguous sample pools, so as to enlarge the total number of samples included.

11. The process of Claim 1 wherein if d is greater than a preset maximum value, d_max, then reducing d to that value.

12. The process of Claim 11, wherein d_max is determined by practical limitations of the method invoked to detect attributes such that desired attributes can be readily detected in a pool with d samples in it so long as d does not exceed d_max.

13. A method of selecting samples, from a set of candidate samples, in accordance with a pattern of desirable attributes wherein sample pools of different biological samples are formed and each attribute in the attribute pattern can be determined by a method capable of detecting, for a particular pool of samples, either the presence of desirable attributes in at least one of the samples, but not all of the samples in that pool, or the presence of desirable attributes in all or none of samples in that pool, the process comprising: determining, for the attributes, the preferred number of samples per pool "d" by setting d as an integer which produces the maximum or a value near the maximum of the product of: d times the number of expected unambiguous sample pools whose constituent samples all have, or do not have, the desirable attributes; dividing the plurality of samples into subsets, comprising approximately d samples per subset such that each of the plurality of samples is included in at least one subset, and, forming unique pools with aliquots from different sample subsets; forming, in the pools, attribute- specific, source-tagged reaction products each having a source tag identifying the pool and a marker-tag identifying a desirable attribute; identifying unambiguous pools wherein all samples in said unambiguous pools have the desirable attributes or no samples do, or identifying ambiguous pools wherein at least one constituent sample does not have the desirable attributes; identifying and selecting samples from unambiguous pools for desirable attributes and either leaving additional attributes in their attribute profiles undetermined , or further analyzing the selected samples; and optionally further analyzing the constituent samples from ambiguous pools to resolve which have and which do not have the desirable attributes in the pattern.

14. The method of claim 13 wherein the attribute pattern reflects the allele or antigen profiles of transfusion-dependent patients.

15. The method of Claim 13 wherein the constituent samples from the ambiguous pools are further analyzed individually to determine the presence of attributes other than desirable attributes.

16. The method of Claim 13 wherein selected samples with the desirable attributes also have unknown attributes which are left undetermined and wherein these attributes are ignored when selecting samples.

17. A method of concurrently determining an attribute profile for two or more biological samples by forming sample pools and determining a first set of attributes for the said pooled samples and as necessary repeating these steps to determine a second set of said attributes, the process comprising:

(i) grouping the attributes in the profile by their known or expected abundances in a population and performing the following steps for each attribute group: selecting an attribute group;

(ii) determining, for the attribute in the selected group with the highest expected abundance, the preferred number "d" of samples per pool by setting d to the integer value which produces the maximum or a value near the maximum of the product of: d times the number of sample pools expected to be unambiguous for said attribute, where a sample pool is unambiguous for an attribute if either all samples in the pool have that attribute or no samples doand otherwise is ambiguous for the attribute;

(iii) dividing the plurality of samples into subsets, comprising approximately d samples per subset such that each of the plurality of samples is included in at least one subset, and, forming unique pools with aliquots from different sample subsets;

(iv) forming, in the pools, attribute- specific reaction products for the constituent samples, each product having a marker-tag identifying a desirable attribute and optionally a source-tag identifying the pool;

(v) identifying samples in pools that are unambiguous for all attributes in the attribute group; and

(vi) identifying samples in pools that are ambiguous for one or more of the attributes in the attribute group and repeating the steps (iii) to (vi) with a smaller value of d.

18. The method of Claim 17 wherein rare attributes are grouped with other rare attributes and more common attributes are grouped with more common attributes.

19. The process of Claim 17 wherein the attributes in the attribute profile are antithetical antigens and wherein for constituent samples of a pool that is ambiguous with respect to antithetical antigens, the presence of at least one sample that is homozygous for a particular antigen is confirmed by an assay wherein a first antibody labeled with a first color is directed to a first of the antithetical antigens, and a second antibody labeled with a second color is directed to a second of the antithetical antigens, to determine the presence or absence, in the pool, of samples expressing both antithetical antigens.

20. The process of Claim 19 wherein the ambiguous pools include samples with both antithetical antigens which are ignored in selecting samples.