METHODS FOR DETERMINING SELECTIVITY OF TEST COMPOUNDS

Abstract

The invention relates to methods for determining the selectivity of a test compound and related methods such as methods for determining whether a subject suffering from cancer will respond or is responsive to treatment with a test compound or compositions comprising more than one test compound.

Claims

1. A method for determining the selectivity of a test compound, the method comprising the steps: (a) providing a sample comprising at least two distinguishable sub-populations of cells in a total population of cells; (b) dividing the sample into at least two parts; (c) incubating at least one part obtained in step (b) in the absence of a test compound and at least one part obtained in step (b) in the presence of a test compound; (d) determining the number of cells in one of the at least two sub-populations that exhibit a distinguishable phenotype, relative to the number of cells in the total population of cells that exhibit the same phenotype in (i) the at least one part incubated in the presence of the test compound and (ii) in the at least one part incubated in the absence of the test compound; and (e) determining selectivity of the test compound to induce the phenotype referred to in (d) in the one sub population referred to in step (d) over all other subpopulations by dividing (i) through (ii) wherein the test compound selectively induces the phenotype referred to in (d) if (i) divided through (ii) is greater than 1, preferably greater than 1.05, 1.1, 1.5, 2, 3 most preferably 5, and selectively inhibits or reduces the phenotype referred to in (d) if it is less than 1, preferably less than 0.95, 0.9, 0.7, 0.5, 0.3, most preferably less than 0.2.

2. The method according to claim 1, wherein the distinguishable phenotype in step (d) is viability and wherein (i) if the selectivity determined in step (e) is <1 the test compound is determined to selectively reduce the number of viable cells of the one sub-population of step (d), and (ii) if the selectivity determined in step (e) is >1 the test compound is determined to selectively improve viability of the one sub-population and/or to selectively reduce the viability of one or more of the sub-population(s) other than the one sub-population of step (d).

3. A method for determining whether a subject suffering from cancer will respond or is responsive to treatment with a test compound, the method comprising the steps: (a) providing a sample obtained from the subject comprising at least two sub-populations of cells in a total population cells, wherein at least one sub-population corresponds to cancerous cells and at least one sub-population corresponds to non-cancerous cells; (b) dividing the sample into at least two parts; (c) incubating at least one part obtained in step (b) in the absence of a test compound and at least one part in the presence of a test compound; (d) determining the number of viable cells in at least one of the sub-populations corresponding to cancer cells relative to the number of viable cells in the total population of cells in (i) the at least one part incubated in the presence of the test compound and (ii) the at least one part incubated in the absence of the test compound; and (e) determining whether the subject will respond or is responsive to treatment with the test compound by dividing (i) through (ii), wherein the subject will respond or is responsive to treatment if the resulting value is less than 1, preferably less than 0.95, 0.9, 0.8, 0.6, 0.4 most preferably less than 0.2.

4. The method of claim 3, wherein the method is repeated for at least two test compounds and whether the subject will respond or is responsive to treatment with a combination of the at least two test compounds is determined by subtracting the values obtained in (e) for each of the at least two test compounds from 1.0, and summing over the resulting values for the at least two test compounds wherein if the resulting sum is greater than −1, preferably greater than −0.5, 0, 0.5, most preferably greater than 1, the subject is determined to respond or be responsive to treatment with the combination of the at least two test compounds.

5. The method according to claim 1, wherein the test compound(s) comprise(s) one or more chemical substances.

6. The method according to claim 1, wherein at least one part obtained in step (b) is further divided into at least two parts, wherein each of the at least two parts is incubated in step (c) with the test compound at different concentrations and wherein steps (d) and (e) are repeated for each concentration of the test compound independently to determine a selectivity/value at each concentration of the test compound whereby an average selectivity/value over all concentrations is calculated after step (e) and used for determining the final selectivity/value.

7. The method according to claim 1, wherein in step (b) the sample is divided into at least three parts and in step (c) at least two parts are incubated in the absence of a test compound and/or at least two parts are incubated in the presence of a test compound whereby each part incubated in the presence of the test compound is incubated in the presence of the same concentration of the test compound, and wherein in step (d) the number of cells in the one of the at least two sub-populations that exhibit a distinguishable phenotype relative to the number of cells in the total population of cells that exhibit the same distinguishable phenotype is determined for (i) each part incubated in the presence of the test compound independently and/or (ii) each part incubated in the absence of the test compound independently and the average of the relative numbers obtained in (i) and/or the average of the relative numbers obtained in (ii) is used.

8. The method according to claim 1, wherein in step (b) the sample is divided into at least three parts and in step (c) at least one part is incubated in the absence of a test compound and/or at least two parts are incubated in the presence of at least two different concentrations of the test compound and in step (d) the number of cells in the one of the at least two sub-populations that exhibit a distinguishable phenotype relative to the number of cells in the total population of cells that exhibit the same distinguishable phenotype is determined for (i) each part incubated in the presence of the test compound independently and/or (ii) each part incubated in the absence of the test compound independently wherein the average of (i) is determined for each concentration independently and/or the average of (ii) is determined and used for further steps and wherein in step (e) the selectivity/value is determined for each concentration of the test compound by dividing the average of (i) for each concentration through the average of (ii) and the final selectivity/value is obtained by averaging the selectivity/value for each concentration.

9. The method according to claim 3, wherein the method is repeated for at least two test compounds and the test compound with the lowest value obtained in step (e) is selected for treatment of the subject suffering from cancer.

10. The method according to claim 4, wherein the method is repeated for at least three test compounds and the combination of at least two of the at least three test compounds with the highest value obtained by subtracting the values obtained in (e) for each of the at least two test compounds in the combination from 1.0, and summing over the resulting values for the at least two test compounds in the combination, is selected for treatment of the subject suffering from cancer.

11. The method according to claim 3, wherein the cancer is a cancer associated with PBMCs or bone marrow cells or cells derived from PBMCs or bone marrow cells.

12. The method according to claim 1, wherein the sample is a tissue sample that contains at least 1% cancerous cells and/or at least 1% non-cancerous cells.

13. The method according to claim 12, wherein the tissue sample is cultured as a non-adherent cell monolayer.

14. The method according to claim 13, wherein the number of viable cancerous and non-cancerous cells is determined using automated microscopy.

15. The method according to claim 14, wherein the number of viable cells is determined as the number of non-fragmented nuclei.

Description

[0109] The present invention is also illustrated in some aspects by the following figures.

[0110] FIG. 1: A. Two hypothetical dose response curves showing viability of cancerous A cells and non-cancerous B cells as a function of drug concentration of a cytotoxic test compound. B. Showing total cell viability and the fraction of viable A cells of viable total cells as a function of test compound concentration. Here, when total cell viability goes below 5% of the starting value, the fraction of viable A cells was set to 0.8 (i.e., the fraction at zero test compound concentration) due to the fact that quantifying small number of cells is associated with large errors. Following this approach, a selectivity of 1 will reliably be assigned to test compounds with strong overall cytotoxicity when following the steps of the methods of the present invention.

[0111] FIG. 2: A. Selectivity/value of a cytotoxic test compounds to kill cancer cells determined as described in the present invention as a function of the difference of log EC50 of the test compound against cancer A cells and non-cancerous B cells measuring at three concentration points (c.f., FIG. 1B). B. Selectivity/value determined as described in the present invention as a function of the difference of log EC50 of the test compound towards cancer cells and non-cancerous cells measuring at 400 concentration points. Already measurement at three concentration points is sufficient to obtain selectivity information using the present invention. However, the more concentration points, the more accurate the selectivity/value reflects the difference in log EC50.

[0112] FIG. 3: Selectivities/values of daunorucibin to kill AML blast cells defined as either being CD34+, CD117+ or CD34+/CD117+ determined for individual daunorubicin concentrations using the present invention displayed as a function of daunorucibin concentration. Patients responding to daunorucibin containing therapy had a lower selectivity/value determined using the present invention than patients not responding to daunorubicin based 3+5+7 induction therapy across different daunorubicin concentrations.

[0113] FIG. 4: Top panel: Selectivity/value of combinations of daunorubicin+cytarabine+etoposide to kill cancer cells (here defined as CD34 or CD117 expressing cells) in bone marrow samples of AML patients determined according the present invention for responders and non-responders to “3+5+7” induction therapy consisting of the three aforementioned drugs. A cut off of 0.92 allows for classification of patients into responders and non-responders with a total classification accuracy of 0.85. Middle panel: Number of cancerous cells relative to number of cancerous cells at zero drug concentration averaged over all drug concentrations and combinations. This metric only allows classification of patients into responders and non-responders with a total classification accuracy of 0.65. Bottom panel: In analogy to middle panel but based on the total cell number. Accurate classification is not possible using this metric.

[0114] FIG. 5: The area under the receiver operator curve was highest (AUROC=0.97) when predicting response of AML patients to daunorubicin containing 3+5+7 induction therapy based on the selectivity of daunorubicin to kill AML blasts determined according to the present invention and using the approach described in Example 4 for downstream data processing. This demonstrates that measuring killing selectivity/value according to the present invention is advantageous compared to e.g., basing the prediction of response on the number of cancerous cells (AUROC=0.91). AML blasts are here defined as in FIG. 2.

[0115] FIG. 6: Left: Hematological cancer patients were treated with combinations of 2 or more FDA approved drugs. For each patient the combined selectivity was calculated as the sum of 1 minus the individual selectivites of the drugs given to the patient as determined according to the present invention. The combined selectivity was plot against the response (PD=progressive disease, SD=stable disease, PR=partial remission, CR=complete remission) and correlates with the response. Based on the integrated selectivity, patients could be classified into responders (CR and PR) and non-responders (PD and SD) with 92% accuracy and an AUROC of 0.84.

[0116] FIG. 7: Selectivities/values of FDA approved drugs to kill CD20+ cells over CD20− cells of a diffuse large B-cell lymphoma patient determined according to the present invention. The patient responded to treatment with ibrutinib.

[0117] FIG. 8: Selectivities/values of FDA approved drugs to kill CD20+ cells over CD20− cells of a B-cell lymphoblastic lymphoma patient determined according to the present invention. The patient responded to treatment with a combination of bortezomib and 6-mercaptopurine.

[0118] FIG. 9: Selectivities/values of FDA approved drugs to kill CD79a+ cells over CD79a− cells of a diffuse large B-cell lymphoma patient determined according to the present invention. The patient responded to treatment with a combination of bortezomib, cladribine and dexamethasone.

[0119] FIG. 10: The viability of cells of population A and population B after treatment with compound X at different concentrations [X] (log EC50 towards A=−2 and log EC50 towards B=3 on an arbitrar concentration scale) was calculated. Accordingly, the number of live cells of population A as a fraction of total live cells (population A+B) was determined as Aviable/(Aviable+Bviable). A logistic dose response curve was fit to the sigmoidal curve resulting from Aviable/(Aviable+Bviable) as a function of the concentration [X] (black line) and the log EC50 determined as the inflection point. Neither log EC50A nor log EC50B correspond to the log EC50 obtained from curve fitting (i.e., the log EC50 of the black line sigmoidal curve) demonstrating that a viability EC50A or EC50B cannot be obtained from fitting a logistic curve to Aviable/(Aviable+Bviable).

EXAMPLES

Example 1

[0120] Synthetic data simulating the response of a mixture of cells comprised of cell population A and B to a cytotoxic drug X was provided. X affected A with a log EC50 of EC50A (e.g., 2.5 for FIG. 1A) and B with a log EC50B (e.g., 3 for FIG. 1B) on an arbitrary concentration scale. Based on these parameters, the number of live cells and total number of cells of each type in the mixture of A and B cells could be calculated assuming a standard 4-parameter logistic (i.e., sigmoidal), dose-response curve. At a concentration of [X]=0, a total number of 10,000 cells at a ratio of A:B=0.8:0.2 was assumed. A measurement of the selectivity of X to kill A over B of a total number of only 3 drug concentrations was simulated.

[0121] Drug selectivity was calculated using the present invention. In particular, for each measured drug concentration the number of viable A cells and the number of viable B cells was calculated. According to step (d) of the methods of the invention (i) the number of viable cells in one of the at least two sub-populations (here A) that exhibit a distinguishable phenotype (here: viability), relative to the number of cells in the total population of cells (here viable A+viable B) that exhibit the same phenotype in the presence of X at three different concentrations as Rx=Ax/(Ax+Bx) where Ax and Bx denote the number of live A and B cells at three different concentrations [X] and (ii) for a concentration of [X]=0 giving R0=A0/(A0+B0) was determined. Then, the selectivity at each concentration of [X] as Sx=Rx/R0 was determined and averaged over all Sx to get the final selectivity of Sfinal=(S1+S2+S3)/3.

[0122] When determining the drug selectivity for different pairs of EC50A and EC50B using the present invention, surprisingly, it was linearly proportional to the difference in log EC50 of X towards A and the log EC50 of X towards B (FIG. 2).

Example 2

[0123] Mononuclear cells were extracted from 20 bone marrow samples of treatment naïve patients newly diagnosed with acute myeloid leukemia (AML) using Ficoll density gradient centrifugation. After bone marrow samples had been taken, all 20 patients had undergone treatment with daunorucibin, etoposide and cytarabine according to the “3+5+7” schedule whereby 10 patients responded and 10 did not.

[0124] The mononuclear cells were suspended in RPMI+10% FCS+penicillin/streptomycin and were seeded into Perkin Elmer Cell Carrier 384-well cell culture plates at a concentration 20,000 cells in 50 μL medium per well whereby the wells had previously been loaded with combinations of cytarabine, daunorubicin and etoposide at different concentrations. All possible drug and concentration combinations were represented on the plate with cytarabine taking concentrations of 0, 1, 3, 10 and 20 μM, daunorubicin taking concentration of 0, 0.1, 1, 3 and 10 μM and etoposide taking concentrations of 0, 1, 3, 10 and 20 μM thus giving a 3-dimension drug titration matrix. Cells were allowed to form monolayers according to WO 2016/046346 and monolayers were incubated for 18 h, fixed by addition of 15 μL 4% formaldehyde solution in PBS containing 0.5% Tritox X114, flicked and stained with DAPI as well fluorescently labelled antibodies to mark CD34 and CD117 positive cells. After 1 h of incubation, images of each well were taken using an Opera Phenix automated confocal microscope (Perkin Elmer).

[0125] Marker positive cells were considered cancerous cells and marker negative cells as non-cancerous cells. The total number of live cells was quantified by counting intact DAPI-stained nuclei using the CellProfiler computational image analysis software whereas fragmented nuclei were discarded as dead or dying. Similarly, the number of live cancerous cells was determined as antibody stained cells with intact DAPI-stained nuclei.

[0126] The selectivity of each drug to kill the cancerous populations was determined according to the present invention by taking the fraction of live cancerous cells of live total cells relative to the fraction of live cancerous cells of live total cells in control wells (no drugs, just DMSO) for different concentrations of daunorubicin only (FIG. 3). Surprisingly, this measure alone could separate responders and non-responders.

[0127] To take into account the contribution of all drugs, the selectivity of each drug to kill the cancerous populations was determined according to the present invention by taking the fraction of live cancerous cells of live total cells at each drug combination and concentration relative to the fraction of live cancerous cells of live total cells in control wells (no drugs, just DMSO) and averaging over all concentrations and combinations of drugs. Using a cut-off value of 0.92, patients who clinically responded to the drug combination could be distinguished from patients who did not (FIG. 4, top panel) with a total classification accuracy of 0.85.

Example 3

[0128] In analogy to Example 2, if drug response was determined only based on the sensitivity of cancer cells (here: CD34 or CD117 positive cells) or the sensitivity of the total cell population, a classification accuracy of 0.65 or less was obtained (FIG. 4, middle and bottom panels, respectively).

Example 4

[0129] In analogy to Examples 2 and 3, this example illustrates how the selectivity determined according to the present invention can be used in downstream analysis to obtain a drug response score that allows for even more accurate classification of patients into responders and non-responders. For each patient sample and combination of drugs at different concentrations, the selectivity was calculated according to the present invention and averaged at each concentration point over responders and non-responders. The resulting data points span dose response surfaces in a four-dimensional dose-response space giving one surface for responders and one surface non-responders. The surface optimally separating the two dose response surfaces was determined by determining the cut-off point at each point in dose-response space that allowed for optimal classification into responders and non-responders at that particular drug dose combination. A response score was calculated for each patient by assigning a 1 to each point in dose response space, that was on the responders' side of the separating surface and a −1 that was on the opposite side. Summing over these indictors weighted by the total classification accuracy at each point in concentration space resulted in a final drug response score. This response score, for example, allowed AML patients receiving 3+5+7 induction therapy to be separated into responders and non-responders with over 90% total classification accuracy (FIG. 5) and area under the receiver operator curve (AUROC) of 0.97 whereas basing the same model on the number of cancerous cells normalized to the number of cancerous cells at zero drug concentration only, gave an AUROC of only 0.91 and basing it on cell number gave an AUROC of 0.86.

Example 5

[0130] Bone marrow aspirates, peripheral blood, pleural effusion, ascites, or excised lymph node samples comprised of cells typically found in PBMCs or bone marrow were purified over Ficoll gradient (bone marrow, peripheral blood, pleural effusion, ascites) (GE healthcare) or homogenized and filtered (lymph tissue) and resuspended in RPMI+10% FCS and penicillin/streptomycin. The resulting single-cell suspensions of mononuclear cells commonly found in PBMCs were seeded in 384-well Perkin Elmer Cell Carrier imaging plates at a concentration of 20,000 cells in 50 μL medium per well plates according to WO 2016/046346 to form non-adherent monolayers. Plates had previously been loaded with 140 different clinically used anticancer drugs in 50 nL DMSO or 50 nL of DMSO as control such that each drug after addition of 50 μL medium and cells was present at 1 or 10 μM final concentration in at least 3 technical replicates per drug and concentration and the DMSO concentration amounted to 0.1% v/v.

[0131] Monolayers were incubated overnight (18 h). Biopsies used for the study were all freshly acquired and not stored frozen. Immunofluorescence staining, imaging by automated microscopy (Opera Phenix, Perkin Elmer), image analysis (CellProfiler), and data analysis (Matlab) were performed as described previously in Vladimer et al Nat Chem Biol 2017. The antibodies used to identify the target cancerous cell populations were selected based on clinical pathology reports and antibody reactivity assessment, and included CD3 (HIT3a), CD19 (HIB19), CD20 (2H7), CD79a (HM47), CD34 (4H11), CD117 (104ED2), and CD138 (DL-101) from eBiosciences. Unstained cells were considered non-cancerous cells.

[0132] The selectivity of drugs to kill cancer cells over non-cancerous cells was determined according to the present invention by taking the average fraction of live cancerous cells of live total cells for each drug and concentration relative to the average fraction of live cancerous cells of live total cells in control wells (no drugs, just DMSO). These quotients of average fractions over the two concentrations for each drug were determined.

[0133] Patients treated with drugs that displayed a value/selectivity of <1 as determined according to the present invention had a higher chance of responding (i.e., achieving a complete or partial remission) than patients treated with drugs that were chosen without taking the value as determined according to the present invention into account or drugs that had a value/selectivity of >1 as determined according to the present invention.

[0134] Moreover, when combinations of drugs were given to patients, the higher the sum of one minus the individual value (FIG. 6) of the drugs given to the patient, the higher the chance that the patient would respond.

Example 6

[0135] A 69-year-old man with Diffuse Large B-cell lymphoma (DLBCL) relapsed after seven lines of prior treatment. Lymphoma cells of the sample were resistant to most of the 104 drugs tested as indicated by a selectivity of >1 of the drugs to kill the cancerous cells relative to non-cancerous cells as determined according the present invention, while only six compounds displayed significant on-target effects ex vivo (FIG. 7). Cisplatin and oxaliplatin were not considered feasible given the patient's history, age, and comorbidities, however the BTK inhibitor ibrutinib showed the second strongest ex vivo efficacy (value/selectivity according to the present invention=0.61, P<0.00048; FIG. 7). A PET-CT performed on day 49 of ibrutinib treatment confirmed a complete remission for the patient.

Example 7

[0136] A 51-year-old women with precursor B-cell lymphoblastic lymphoma (B-LBL) had undergone three lines of prior treatment, and was progressive after immunotherapy with the bi-specific CD3-CD19 antibody blinatumomab. A cell mixture comprising cells typically found in PBMCs was isolated from, the woman's pleural effusion. The ability of 266 compounds to selectively kill cancerous versus non-cancerous cells contained in the cell mixture was determined using the present invention. It revealed that the proteasome inhibitor bortezomib was able to selectively kill cancer cancer cells (selectivity determined according to the present invention=0.50, P<0.001; FIG. 8) and the thiopurine 6-mercaptopurine, 6-MP (value/selectivity determined according to the present invention=0.58, P<0.001; FIG. 8). 6-MP and bortezomib were combined with anti-CD20 obinutuzumab. After 28 days PET-CT confirmed a partial response.

Example 8

[0137] An excised lymph node of a patient with diffuse large B-cell lymphoma was dissociated into single cells giving a complex cell mixture comprising cells typically found in PBMCs. The ability of 266 compounds to selectively kill cancerous versus non-cancerous cells contained in the cell mixture according to the present invention. The patient achieved a complete remission (FIG. 9) to a combination of the single strongest ex vivo acting drug, bortezomib (selectivity=0.59, P<0.0001;), with cladribine (selectivity=0.73; P<0.0003) and dexamethasone (value/selectivity=0.87; P<0.05; FIG. 9).

Example 9

[0138] This example describes the practical application of the methods described in claims (1) and following and in particular of claims 1) and (2). A tissue sample of a blood cancer patient comprised of 40,000 cells is provided. 20,000 cells are cancerous cells and stain positive for the cell surface marker CD19. The remaining cells stain positive for other cell surface markers including CD3, 4, 8, 11c, 14, 56 and others. The sample is divided into two parts of 20,000 cells each. The first sample is incubated in RPMI+10% FCS in the presence of 10 μM bortezomib in DMSO (0.1% final DMSO concentration) whereas the second sample is incubated in RPMI+10% FCS+0.1% DMSO. After 24 h incubation the viability of each cell in each sample is determined whereby viability here is the “distinguishable phenotype” referred to in step (d) of claim 1 and dependent claims. In the bortezomib treated sample, 5,000 viable cells staining for CD19 are found and 10,000 viable cells negative for CD19 stain are found. In the DMSO treated sample 10,000 viable cells staining positive for CD19 and staining negative for CD19 are found each. According to the present invention, the selectivity of bortezomib to reduce viability of the CD19 positive cells is calculated. Following step (d) the number of cells in one of the at least two sub-populations (here: CD19 positive cells) that exhibit a distinguishable phenotype (here: viability), relative to the number of cells in the total population of cells (CD19 positive+CD19 negative cells) that exhibit the same phenotype (i.e., viability) in (i) the at least one part incubated in the presence of bortezomib as the test compound (here: 5,000/15,000=0.33) and (ii) in the at least one part incubated in the absence of the test compound (here: 10,000/20,000=0.5) is calculated.

[0139] Following step (e) the selectivity of the test compound (here: bortezomib) to induce the phenotype referred to in (d) in the one sub population referred to in step (d) (here: CD19 positive cells) over all other subpopulations is determined by dividing (i) (here: 0.33) through (ii) (here: 0.50). Since 0.33/0.50=0.66 i.e., less than 1, the test compound (here: bortezomib) selectively inhibits the phenotype of steps (d) (here: viability) in the one population explicitly referred to in step (d) (here: CD19 positive cells). So, according to the present invention we can conclude that bortezomib selectively reduced the viability of CD19 positive cells in the given example.

Example 10

[0140] This example describes a further practical application of the methods of the invention. A tissue sample of a blood cancer patient comprised of 60,000 cells is provided. 30,000 cells are cancerous cells and stain positive for the cell surface marker CD79a. The remaining cells stain positive for other cell surface markers including CD3, 4, 8, 11c, 14, 56 and others. Suitably labelled antibodies are used as staining reagents.

[0141] The sample is divided into two parts of 40,000 cells and 20,000 cells. The first part of 40,000 cells is further divided into two parts of 20,000 cells each here denoted [1a] and [1b]. Parts [1a] and [1b] are incubated in RPMI+10% FCS in the presence of 10 μM and 1 μM bortezomib in DMSO (0.1% final DMSO concentration) respectively whereas the second sample is incubated in RPMI+10% FCS+0.1% DMSO. CD79a here is only chosen as a hypothetical example for the sake of clarity and can be replaced with any other surface marker.

[0142] After 24 h incubation the viability of each cell in each sample is determined whereby viability here is the “distinguishable phenotype” as used in the present invention. In the bortezomib treated sample [1a], 5,000 viable cells staining for CD79a are found and 10,000 viable cells negative for CD79a stain are found. In the bortezomib treated sample [1b], 8,000 viable cells staining for CD79a are found and 10,000 viable cells negative for CD79a stain are found. In the DMSO treated sample 10,000 viable cells staining positive for CD79a and staining negative for CD79a are found each. The selectivity of bortezomib to reduce viability of the CD79a positive cells is calculated according to the methods of the invention.

[0143] For both parts [1a] and [1b] the number of, cells in the one of the at least two sub-populations (here: CD79a positive cells) that exhibit a distinguishable phenotype (here: viability), relative to the number of cells in the total population of cells (CD79a positive+CD79a negative cells) that exhibit the same phenotype (i.e., viability) in (i) the at least one part incubated in the presence of bortezomib at the respective concentration as the test compound (here: 5,000/15,000=0.33 for [1a] and 8,000/18,000=0.44 for [1b]) and (ii) in the at least one part incubated in the absence of the test compound (here: 10,000/20,000=0.5) is calculated.

[0144] Following step (e) the selectivity of the test compound (here: bortezomib) to induce the phenotype referred to in (d) in the one sub-population referred to in step (d) (here: CD79a positive cells) over all other subpopulations is determined by dividing (i) (here: 0.33 for [1a] and 0.44 for [1b]) through (ii) (here: 0.50) and the average selectivity is calculated as the final value/selectivity as (0.33/0.50+0.44/0.50)/2=0.77. Since 0.77 is less than 1, the test compound (here: bortezomib) selectively inhibits the phenotype of steps (d) (here: viability) in the one population explicitly referred to in step (d) (here: CD79 positive cells). So, according to the present invention it can be concluded that bortezomib selectively reduced the viability of CD79a positive cells in the given example.

[0145] The patient from wham the sample was derived would respond to treatment with bortezomib. CD79a here is only chosen as a hypothetical example for the sake of clarity and can be replaced with any other surface marker. Also cell numbers are only chosen arbitrarily for illustrative purposes.

Example 11

[0146] A tissue sample of a blood cancer patient comprised of 60,000 cells is provided. 30,000 cells are cancerous cells and stain positive for the cell surface marker CD20. The remaining cells stain positive for other cell surface markers including CD3, 4, 8, 11c, 14, 56 and others. Suitably labelled antibodies are used as staining reagents. The sample is divided into three parts of 20,000 cells each. Two parts of 20,000 each are incubated in RPMI+10% FCS in the presence of 10 μM bortezomib in DMSO (0.1% final DMSO concentration) respectively whereas the third part is incubated in RPMI+10% FCS+0.1% DMSO. Please note that CD20a here is only chosen as a hypothetical example for the sake of clarity and can be replaced with any other surface marker.

[0147] After 24 h incubation the viability of each cell in each sample is determined whereby viability here is the “distinguishable phenotype”. In the two samples treated with 10 μM bortezomib, 5,000 viable cells staining for CD20a are found and 10,000 viable cells negative for CD20a stain are found each. In the DMSO treated sample 10,000 viable cells staining positive for CD20 and staining negative for CD20 are found each. The selectivity of bortezomib to reduce viability of the CD79a positive cells is determined. For both parts incubated in the presence of 10 μM bortezomib the number of cells in the one of the at least two sub-populations, (here: CD20 positive cells) that exhibit a distinguishable phenotype (here: viability), relative to the number of cells in the total population of cells (CD20 positive+CD20 negative cells) that exhibit the same phenotype (i.e., viability) in (i) each part incubated in the presence of bortezomib is calculated independently (i.e., 5,000/15,000=0.33 and 5,000/15,000=0.33) and (ii) for each part incubated in the absence of the test compound (here: 10,000/20,000=0.5) is determined independently. Then the average of (i) and (ii) are formed giving 0.33 for (i) and 0.5 for (ii) and used for further steps, that is, in step (e) the value/selectivity is determined by dividing the average of (i) by the average of (ii) giving 0.33/0.5=0.66 as the final value/selectivity.

Example 12

[0148] This example illustrates that accurate EC50 values cannot be obtained from fitting dose response curves to fractions of cells exhibiting a phenotype of total cells exhibiting the same phenotype. A mixture of cells of type A and B at a ratio of A:B=0.2:0.8 was assumed. The cell mixture was treated with a cytotoxic compound X. The ability of compound X to kill A cells was quantified with a log EC50 of 3 and the ability of compound X to kill B cells was quantified with a log EC50 of −2 on an arbitrary concentration scale. Calculating the fraction of the number live A cells of the total number of live cells (i.e. live A+live B), the sigmoidal curve shown in FIG. 10 was obtained. It can be clearly seen, that the inflection point of this curve (solid line) is neither informative of the EC50 of the dose response curve of X action on A nor of X action on B.

Example 13

[0149] This example illustrates the effect of a 10% standard deviation in total cell numbers when introducing a cell mixture of A and B cells into a microtiter plate for determining selectivity of a test compound X to kill A over B cells. When determining the selectivity using the classic approach of measuring total number of A and B cells as a function of concentration [X], to fit sigmoidal dose response curves and measure the EC50 of X towards A and B, each measurement point will have a standard deviation of 10%. Using the present invention, a 10% variation in total cell number will have no effect on the fraction of viable A cells of the total number of viable cells. The present invention thus allows for the determination of selectivity that is more robust towards variation in seeding of cells into assay plates or loss of cells during manipulation.