METHOD OF PROFILING A SAMPLE COMPRISING A PLURALITY OF CELLS AND A SYSTEM FOR PERFORMING THE SAME

Abstract

The invention is to provide a method of profiling a sample comprising a plurality of cells, the method comprising: flowing cells from the sample through a first array of pillars to obtain one or more distribution profiles of cells sorted by the first array; flowing cells from the sample through a second array of pillars that is different from the first array of pillars to obtain on one or more distribution profiles of cells sorted by the second array; and deriving a biophysical signature of the sample based on at least the one or more distribution profiles of the cells sorted by the first array and/or the one or more distribution profiles of the cells sorted by the second array. The method further comprises determining a health status of a subject based on the biophysical signature of the sample. The invention is also to provide a sample profiling system. In various embodiments, the distribution profile of cells in the output regions is indicative of one or more biophysical properties of the cells, which may include the size and deformability of the cells. The pillars in the first array and the second array may have a shape selected from the group consisting of a substantially L shape and a substantially inverse L shape, mirror reflections thereof or combinations thereof.

Claims

1. A method of profiling a sample comprising a plurality of cells, the method comprising: flowing cells from the sample through a first array of pillars to obtain one or more distribution profiles of cells sorted by the first array; flowing cells from the sample through a second array of pillars that is different from the first array of pillars to obtain on one or more distribution profiles of cells sorted by the second array; and deriving a biophysical signature of the sample based on at least the one or more distribution profiles of the cells sorted by the first array and/or the one or more distribution profiles of the cells sorted by the second array.

2. The method of claim 1, wherein flowing cells through the first array of pillars comprises flowing the cells through the first array of pillars at different flow velocities and flowing cells through the second array of pillars comprises flowing the cells through the second array of pillars at different flow velocities or flow rates.

3. The method of claim 1, further comprising obtaining a first biophysical parameter based on the one or more distribution profiles of the cells sorted by the first array and/or obtaining a second biophysical parameter based on one or more distribution profiles of the cells sorted by the second array.

4. The method of claim 3, wherein obtaining the first biophysical parameter and/or second biophysical parameter comprises determining a cell apparent size (D.sub.app) based on the one or more distribution profiles of the sorted cells, optionally determining respective cell apparent sizes (D.sub.app) based on the respective distribution profiles of the sorted cells at the respective different flow velocities or flow rates.

5. The method of claim 4, wherein obtaining the first biophysical parameter and/or the second biophysical parameter further comprises obtaining a cell-deformability modulus (CDM), optionally based on changes in the cell apparent sizes (D.sub.app) at different flow velocities or flow rates.

6. The method of claim 5, wherein the biophysical signature of the sample is derived from the respective cell-deformability modulus (CDM) obtained for at least the first array of pillars and the second array of pillars.

7. The method of claim 1, wherein the pillars of each the first and second arrays are arranged based on equation (A):
Dc=ag tan θ.sup.b  (A) where D.sub.c is the deterministic lateral displacement (DLD) cut-off size, each of a and b is a value that is independently selected from a value in the range of 0.48 to 1.4 and g represents the closest distance between the pillars.

8. The method of claim 7, wherein D.sub.c is in the range of 5.0 μm to 16.0 μm.

9. The method of claim 1, wherein the first array of pillars differs from the second array of pillars in at least one of: pillar dimension, pillar shape, pillar structure, pillar arrangement or pillar orientation, with respect to the direction of flow of cells.

10. The method of claim 1, wherein the pillars in the first array and the second array have a shape selected from the group consisting of a substantially L shape (L), a substantially inverse L shape (L.sup.−1), mirror reflections thereof or combinations thereof.

11. The method of claim 1, wherein the sample is derived from a mammalian subject and the method further comprises determining a health status of a subject based on the biophysical signature of the sample.

12. The method of claim 11, wherein determining a health status of a subject comprises determining the presence of an infection and/or inflammation in the subject.

13. The method of claim 12, wherein the cells comprise immune cells.

14. A sample profiling system comprising: a first region comprising a first array of pillars configured to sort cells from a sample flowed therethrough and provide one or more distribution profiles of the sorted cells; and a second region comprising a second array of pillars configured to sort cells from the sample flowed therethrough and provide one or more distribution profiles of the sorted cells; wherein the first array of pillars is configured to provide one or more distribution profiles that is substantially different from the one or more distribution profiles provided by the second array of pillars for the same sample.

15. The system of claim 14, wherein each of the first and second regions is fluidically coupled to at least one input reservoir and at least one output port.

16. The system of claim 14, wherein the pillars of each the first and second array are arranged based on equation (A):
Dc=ag tan θ.sup.b  (A) where D.sub.c is the deterministic lateral displacement (DLD) cut-off size, each of a and b is a value that is independently selected from a value in the range of 0.48 to 1.4 and g represents the closest distance between the pillars.

17. The system of claim 16, wherein the first region comprising the first array of pillars and the second region comprising the second array of pillars each comprise a plurality of segments, each segment differing from the adjacent segment by the offsetting angle of the pillars (θ) and the corresponding DLD cut-off size (D.sub.c).

18. The system of claim 16, wherein D.sub.c is in the range of 5.0 μm to 16.0 μm.

19. The system of claim 14, wherein the first array of pillars differs from the second array of pillars in at least one of: pillar dimension, pillar shape, pillar structure, pillar arrangement or pillar orientation, with reference to the direction of flow of cells.

20. The system of claim 14, wherein the system further comprises at least one detection setup for obtaining the one or more distribution profiles of the cells sorted by the first array and/or second array.

Description

BRIEF DESCRIPTION OF FIGURES

[0082] FIGS. 1A and 1B are schematic drawings illustrating a DLD device used for immune cell profiling assay and an immune profiling workflow using DLD assays for L and L.sup.−1 pillar shapes, respectively, in accordance with various embodiments disclosed herein. FIG. 1A shows the whole blood DLD assay by loading the blood into the sample reservoir 102A of the PDMS DLD device (or system) 100 which is used to simultaneously sort and measure the distribution of cells across the output region allowing size frequency distribution analysis. The device 100 comprises of two additional buffer reservoirs 102B and 102C which sandwich the sample stream resulting in a precise injection of sample into the DLD region. The DLD region is composed of 21 DLD segments corresponding to 21 step measurement resolution ranging from size 6.0 to 16.0 μm in steps of 0.5 μm. Scale bar is 200 μm. FIG. 1B shows the DLD assay(s) used to profile WBC based on their unique biophysical signatures in the different DLD pillar structures. These biophysical parameters are used to then classify the immune spectrum from healthy to severe immune response.

[0083] FIGS. 2A and 2B are schematic drawings showing the specifications for DLD devices 1 and 2, respectively, in accordance with various embodiments disclosed herein. θ.sub.seg changes depending on the 21 DLD segments.

[0084] FIGS. 3A, 3B and 3C are graphs showing size and deformability measurements of WBCs in L and L.sup.−1 DLD devices in accordance with various embodiments disclosed herein. The frequency distribution plot for the measure D.sub.app of WBCs at various flow velocities in different devices are shown in FIG. 3A for L and in FIG. 3B for L.sup.−1. The L ΔD.sub.app and L.sup.−1 ΔD.sub.app were measured at 2.0 and 3.0 μm, respectively. n>100 were used for each distribution and the error bar denotes the sample standard deviation. FIG. 3C introduces the cell DLD-deformability modulus (CDM) parameter where the rate of change of size can be measured by plotting the fitting equations of the size plots for L and L.sup.−1.

[0085] FIGS. 4A and 4B are graphs illustrating DLD device characterisation using bead standards at different flow velocities in accordance with various embodiments disclosed herein. FIG. 4A shows a graph plot of measure apparent size, D.sub.app, versus size of beads at a flow of 2.5 μL/min. Four size standards of 6.2, 7.3, 8.2 and 10.2 μm beads were used to calibrate the devices. Ideally, if the designed specifications of DLD fits perfectly, D.sub.app will be equivalent to the size of the beads as depicted in the dotted line. The top half triangular region depicts a condition where D.sub.app>size of beads and the converse is true for the bottom half triangular region. L and L.sup.−1 sorting performance of beads are measured in the respective plots. n>200 for each point and error bar is S.D. of sample population. FIG. 4B shows the measurement of mean D.sub.app size of beads at various flow velocities. The mean of mean plot shows the average D.sub.app for beads flow at various velocities (n=4) and error bar denotes the S.D. of these means.

[0086] FIGS. 5A, 5B, 5C and 5D are images illustrating WBC paths over L.sup.−1 and L pillars in accordance with various embodiments disclosed herein. FIGS. 5A and 5B show the instantaneous simulated flow streamlines around DLD pillars. Magnified experimental time-lapse overlay of individual WBC trajectories and dynamics over L and L.sup.−1 structures is seen in FIGS. 5C and 5D, respectively. The darker outline and lighter outline are pseudo-shades highlighting experimental WBC sequential motion at two flow rates, Q.sub.1=2.5 μL/min and Q.sub.2=25.0 μL/min, respectively.

[0087] FIG. 6 is an image showing WBC overlay images for L and L.sup.−1 DLD structures in accordance with various embodiments disclosed herein.

[0088] FIGS. 7A, 7B, 7C, 7D and 7E are graphs and schematic drawings on the measurements performed for and results from DLD assays for WBC biophysical measurements of size and deformability from whole blood in accordance with various embodiments disclosed herein. n=5 healthy donor samples were used to measure the size parameters in FIG. 7A and deformation parameters in FIG. 7B. D.sub.app were measured at 2.5 μL/min for L and L.sup.−1 and a paired t-test with ** denoting p=0.004. Average D.sub.app is the mean of both L and L.sup.−1 where n=5 and error bar denotes standard deviation of sample. The CDM deformation parameter for L and L.sup.−1 were plotted in FIG. 7B with *** denoting a p<0.001 for a paired t-test of n=5 sample. CDM.sub.dot is the product of both CDMA and CDM.sub.L-1. FIG. 7C shows three groups of measurements performed, namely biophysical profiling of WBCs from direct sample injection, in vitro WBC assays and common blood processing/storage methods. Direct sample injections include a healthy donor control sample, emergency department (ED) admission control (i.e. patients with no clear signs of infection) and ED admission with infection and two or more systemic inflammatory response syndrome (SIRS) criteria. The mean size measurement, D.sub.app, across all samples are depicted in FIG. 7D while the deformability parameter CDM.sub.dot is shown in FIG. 7E. The horizontal dotted line denotes the mean value of healthy donor in FIGS. 7D and 7E with standard deviation shown in the shaded region. n=5 samples were used for all plots with standard deviation represented by the error bar.

[0089] FIG. 8 is a graph showing a comparison of various CDM measurements of CDM.sub.L, CDM.sub.L-1 and CDM.sub.dot in accordance with various embodiments disclosed herein.

[0090] FIG. 9 is a graph showing a 38 biophysical marker Principal Component Analysis (PCA) plot in accordance with various embodiments disclosed herein.

[0091] FIGS. 10A, 10B, 10C, 10D, 10E, 10F, 10G and 10H are graphs and an image comparing label-free biophysical immune markers and signatures of various immune status in accordance with various embodiments disclosed herein. FIGS. 10A to 10D show plots for Size 1 to Size 4 features, while FIGS. 10E to 10H show plots for cell Count 1-Count 4, respectively from a list of 38 biophysical markers (see Table 5). The plots compare the mean and sample standard deviation for healthy (n=8) samples, Control (n=36) samples and ≥2 SIRS (n=41) samples. An independent two tailed t-test is used to compute the p-values of the sample measurements with n.s. denoting not significant, * for p<0.05, ** for p<0.01, *** for p<0.001 and **** for p<0.0001. FIG. 10I shows the hierarchical clustering and heatmap of normalized biophysical marker value of all 85 samples comprising healthy, no infection control, infection tests >2 SIRS and severe immune response with >2 SIRS. 8 clusters were identified based on the data and the heatmap shows the corresponding biomarker signatures. The biomarkers are grouped based on size, deformability, distribution and cell count.

[0092] FIG. 11 is an image showing a 38 biomarker features correlation heatmap in accordance with various embodiments disclosed herein.

[0093] FIG. 12 is an image showing a comparison of hierarchical clustering of 38 biomarker signatures upon admission to ED and hospitalization stay in accordance with various embodiments disclosed herein.

[0094] FIG. 13 is a graph showing a ROC curve plotting the True Positive Rate against the False Positive Rate with the area under curve (AUC) at 0.97 in accordance with various embodiments disclosed herein.

[0095] FIG. 14 is a flowchart showing an algorithm for the ROC plotting of all data features for non-infection vs infection controls and classification metrics calculation using the SVM classifier model in accordance with various embodiments disclosed herein.

[0096] FIG. 15 is a schematic drawing illustrating a system comprising a DLD device in an exemplary embodiment.

EXAMPLES

[0097] Example embodiments of the disclosure will be better understood and readily apparent to one of ordinary skill in the art from the following discussions and if applicable, in conjunction with the figures. It should be appreciated that other modifications related to biological, chemical, structural, electrical and optical changes may be made without deviating from the scope of the invention. Example embodiments are not necessarily mutually exclusive as some may be combined with one or more embodiments to form new exemplary embodiments.

[0098] Disease manifestation and severity from acute infections are often due to hyper-aggressive host immune responses which changes within minutes. Current methods for early diagnosis of infections focus on detecting low abundance pathogens, which are time-consuming, of low sensitivity, and does not reflect the severity of the pathophysiology appropriately.

[0099] The examples describe a rapid label-free immune profiling deterministic lateral displacement (DLD) assay as a quantitative diagnostic measure of immune cell biophysical signature using 20 μL of whole undiluted and unprocessed blood in under 15 minutes. The approach here focuses on profiling the rapidly changing host inflammatory response, which in its over-exuberant state, leads to sepsis and death. In embodiments disclosed herein, the assay is based on a simple workflow where whole blood is loaded onto a microfluidic chip (or a system) and the DLD assay simultaneously sort immune cells (WBC) from whole blood and profile the biophysical properties of size, deformation, distribution and cell count which correlates to the immune states. The deterministic nature of particle interactions within DLD devices result in predictable and high-resolution (˜10 nm) sorting. As will be shown in the following examples, unconventional L and inverse-L (L.sup.−1) DLD pillar structures interact and sort WBCs differently resulting in unique biophysical signatures. DLD precision sorting was translated into an assay to quantify and profile the immune states of WBCs reflecting severity of immune response. The hydrodynamic interactions of deformable immune cells enable simultaneous sorting and immune response profiling in whole blood.

[0100] In the following examples, the biophysical DLD assay was performed directly on whole blood samples from healthy donors and patients recruited from the ED. Interestingly, the DLD assay reveals divergent biophysical signatures of immune cells from patients with infection versus immune cells triggered in vitro with known activators such as lipopolysaccharides (LPS) and phorbol 12-myristate 13-acetate (PMA). These findings suggest in vitro immune cell activation do not mimic physiological immune cell response and emphasize the significance of this work on profiling immune cells in its native physiological state—whole blood with minimal perturbation.

[0101] In the following examples, the diagnostic modality was evaluated by recruiting 8 healthy donors, 36 donors with non-infection symptoms such as cardiac conditions and 41 donors presenting to the ED with 2 or more components of the systemic inflammatory response syndrome (SIRS). The DLD assay on a single drop of blood reveals significant immune biophysical response signatures which resulted in distinction between infection and non-infection group with a detection sensitivity of 0.91 and specificity of 0.92.

[0102] In the following examples, with a whole blood sample throughput of up to 10,000 cells/s using video captured frame rates of 15 to 150 frames per second (fps), it is shown that the biophysical diagnostic modality can be easily achieved using low-cost and compact machine vision cameras or smart phone optical sensors making it attractive for deployable point-of-care systems for rapid patient triage of immune dysregulation in ED. This could potentially change disease diagnosis, treatment, and risk management in the settings of primary care and hospitals.

[0103] The preliminary clinical study of the 85 donors in emergency department with a spectrum of immune response states from healthy to severe inflammatory response shows correlation with biophysical markers of immune cell size, deformability, distribution, and cell counts. The speed of patient stratification demonstrated here has promising impact in deployable point-of-care systems for acute infections triage, risk management and resource allocation at emergency departments, where clinical manifestation of infections severity may not be clinically evident as compared to inpatients in the wards or intensive care units.

[0104] WBC Biophysical Measurements in DLD Device

[0105] FIG. 1A shows DLD devices (or systems) 100 used for immune cell profiling assay consist of a polydimethylsiloxane (PDMS) device with three open reservoirs with respective inlet ports 102A, 102B and 102C, and a single outlet tubing coupled to an outlet port 104 and attached to a syringe pump. The open reservoirs 102A, 102B and 102C facilitate easy sample loading, sample resuspension to prevent settling of cells and washing the reservoir to reuse the device. The required loaded volume per run is 10 μL and the reservoir can hold up to 25 μL of blood. As the fluid is withdrawn, the sample flows through a region comprising 21 DLD device segments sandwiched between two 1× phosphate-buffered saline (PBS) buffer streams (see Table 2). Each segment comprises an array of pillars and has a specific DLD critical cut-off size (D.sub.c) determined by the empirical Equation (1):

D.sub.c=1.4 G tan θ.sup.0.48  (1)

where G is the regular spacing between pillars and θ is the gradient of the pillar array. This design is known as a chirped DLD array where each downstream segment has an increasing pillar row-shift gradient corresponding to an increasing D.sub.c ranging from 6.0 to 16.0 μm in steps of 0.5 μm (see Methods). Immune cells flowing through the device 100 are deflected laterally only within DLD segments where cell sizes are larger than D.sub.c; the cells therefore exit the device 100 at defined lateral positions depicted in the output region shown in FIG. 1A. The output of the sorting forms a spectrum in its size distribution (i.e., a biophysical parameter based on a distribution profile of the cells sorted by the array of pillars). The apparent cell size (D.sub.app) is the size that is exhibited in a DLD microfluidic device given the design parameters D.sub.c from Equation (1) and the observed outlet distribution.

TABLE-US-00001 TABLE 2 DLD segments parameters for D.sub.c calculation based the DLD pillar dimensions shown in FIGs. 2A and 2B. Dc Gradient Segment Segment (μm) N (Deg) Length (μm)  1 6.0 33.13 1.73 1656  2 6.5 28.04 2.04 1402  3 7.0 24.03 2.38 1201  4 7.5 20.81 2.75 1041  5 8.0 18.19 3.15 910  6 8.5 16.04 3.57 802  7 9.0 14.24 4.02 712  8 9.5 12.72 4.50 636  9 10.0 11.43 5.00 571 10 10.5 10.32 5.53 516 11 11.0 9.37 6.09 469 12 11.5 8.54 6.68 427 13 12.0 7.82 7.29 391 14 12.5 7.18 7.93 359 15 13.0 6.62 8.59 331 16 13.5 6.12 9.29 306 17 14.0 5.67 10.00 284 18 14.5 5.27 10.74 264 19 15.0 4.91 11.51 246 20 15.5 4.59 12.30 229 21 16.0 4.29 13.11 215

[0106] The DLD assay has a minimum measurable D.sub.app of 6.0 μm, and RBCs having an apparent size of less than 3.0 μm would not be deflected laterally in the DLD device. As such, the input and output lateral position of RBCs remains the same, albeit with a larger spread at the outlet region. This spread is due to diffusive effects and the stochastic nature of RBC interaction within the DLD (compare images of input region and output region shown in FIG. 1A). The distribution of WBCs across the outlet can be counted and analysed for its apparent mean size and standard deviation (S.D.).

[0107] Two DLD pillar structures were investigated in this example, namely L and L.sup.−1 (see FIG. 1B and FIGS. 2A and 2B). Previous studies have shown contrasting sorting effects of these two pillars on the highly deformable and biconcave disc-shaped RBC. Despite the preliminary evidence of size and shape deformability sorting of RBCs, information on DLD pillar shape effects on generally spherical and deformable WBCs is lacking. In this example, the unique WBC sorting signatures (or biophysical signatures) of these different DLD pillar structures are utilized as an assay to profile the activation state of WBCs (see FIG. 1B). By using different flow velocities, each DLD assay elicits a unique biophysical interaction with deformable WBCs. These biophysical traits and parameters are aggregated and used to classify the WBC state as activated or non-activated.

[0108] Effect of Flow Rates on WBC Size and Deformation

[0109] WBCs are deformable particles and their morphology changes with application of external forces. As shown in FIGS. 3A and 3B, the D.sub.app of WBCs decreases as fluid flow rate increases. For the L-shaped DLD assay, the WBC output spectrum shows a mean D.sub.app of 9.7 μm, 9.3 μm, 8.2 μm and 7.7 μm for flow rates of 2.5, 5.0, 10.0 and 25.0 μL/min, respectively (see FIG. 3A). In the L.sup.−1 DLD device, WBCs have mean D.sub.app from 10.1 μm, 9.5 μm, 8.6 μm to 7.1 μm (see FIG. 3B). The reduction in mean D.sub.app at 2.5 μL/min to 25 μL/min is denoted by ΔD.sub.app, which corresponds to L ΔD.sub.app=2.0 μm and L.sup.−1 ΔD.sub.app=3.0 μm. This measures up to at least 15% reduction in mean D.sub.app as the flow rate increases.

[0110] The difference between two DLD assays using the same sample can be interpreted clearer in the graph plot shown in FIG. 3C where D.sub.app is plotted against flow velocity. The trend is linear in the logarithmic scale, resulting in a log-linear equation measuring the change of D.sub.app with respect to fluid flow velocity. In FIG. 3C, the modulus of the gradient is defined as DLD cell-deformability modulus (CDM). The CDM parameter quantifies the change in WBC apparent size over varying flow velocities from the measurement at 2.5 μL/min. The CDM parameter for L and L.sup.−1 are denoted as CDM.sub.L and CDM.sub.L-1, respectively. As the L DLD assay showed a smaller L ΔD.sub.app (see FIG. 3A), the CDM.sub.L in FIG. 3C is correspondingly smaller at 0.94 as compared to CDM.sub.L-1 at 1.31.

[0111] As shown in FIGS. 3A to 3C, DLD profiling assays using L and L.sup.−1 pillar shapes lead to different sorting signatures for deformable cells. On the contrary, the same DLD assay performed on rigid beads shows no significant change in D.sub.app at different flow rates for L and L.sup.−1 DLD tests (see Supplementary Discussion 1 below and FIGS. 4A and 4B). This example strongly demonstrates that unequal cell deformation results in different flow outcomes in L and L.sup.−1 DLD pillar structures.

[0112] Supplementary Discussion 1: Device Characterisation and Flow Performance

[0113] Beads of 6.2, 7.2, 8.3 and 10.2 μm sizes were used for characterisation of both L and L.sup.−1 DLD devices. FIG. 4A shows the characterized plots of size of beads versus the measured D.sub.app of various beads in L and L.sup.−1 DLD devices. The flow rate used is 2.5 μL/min. The boundary demarcating the top half and bottom half triangular region is the theoretical boundary for which the measured D.sub.app is equivalent to designed specifications of D.sub.c based on Equation (1). Points in the top half triangular region above the central line denote increased DLD sorting performance where a change in size of beads result in a larger change in measured D.sub.app. Both L and L.sup.−1 characterized sorting plots lies within the top half triangular region, because the beads are not deformable at all.

[0114] This enhancement in D.sub.app is not unexpected. It is noted that L and L.sup.−1 structures constitute a class of DLD structures known to induce asymmetric fluid flow profiles which increases the sorting effectiveness relative to symmetric flow profiles of circle pillar structures. This implies that for the same DLD gap and angle, a smaller specific D.sub.c can be achieved. However, what is assumed here is the skew and linear relationship based on the dotted line plot in FIG. 4A. Since this is a chirped DLD design, small incremental pillar shape enhancement in each DLD segments adds up, which results in the corresponding skew.

[0115] Two interesting observations are highlighted here. Firstly, the linear plot skew represents a 1.5× amplification of bead size measurement for L and L.sup.−1 DLD pillars (see FIG. 4A). A 1.0 μm change in bead size will result in a 1.5 μm difference in measurement. This factor is vital as measurements in deformability is simply the change in size over various flow velocities and thus a 1.5× amplification measurement suggest increased sensitivity over the designed specifications based on D.sub.c=D.sub.app=Size. Secondly, the S.D. of the measure D.sub.app is the same or smaller than the S.D. provided by the manufacturer. This suggests DLD measurements of D.sub.app can accurately quantify the size and S.D. of the beads.

[0116] The effects of fluid flow velocities on sorting of rigid spherical beads were evaluated in FIG. 4B and the raw histogram with mean and S.D. of the various measurements are shown. Fluid velocities varied from 2.5 μL/min to 25.0 μL/min for all the beads in the two DLD pillars. Altering flow rates of beads seem to vary within 0.5 μm. Thus, the impact of increasing flow velocities minimally changes D.sub.app suggesting that rigid beads do not change its D.sub.app size in these flow rates with minimal deformation. Since Reynolds number in this setting is Re<1 (˜0.05-0.5), laminar flow is also expected resulting in a deterministic and predictable behaviour of rigid beads.

[0117] Visualizing WBC Flow Signatures in DLD Assays

[0118] In the exemplary DLD assays, measured D.sub.app varies depending on the pillar structure. This is primarily due to WBC deformability, resulting in differences in their periodic flow trajectories as they navigate between the two consecutive DLD pillar micro-structures. The simulated hydrodynamic streamlines visualize the fluid motion with respect to the cell (see FIGS. 5A and 5B), clearly showing fluid flow differences. This principle becomes clearer when the experimental cell trajectory is tracked within a small DLD pillar unit for L and L.sup.−1 structure (see FIGS. 5C and 5D). Each pillar structure interacts with the deformable cell in complex ways, and it is evident that the paths taken by cells at slow (Q.sub.1=2.5 μL/min) and fast (Q.sub.2=25.0 μL/min) flow rates differ between L and L.sup.−1 DLD pillars. DLD relies on repetitive interactions between stationary pillars and moving cells, any small difference in cell path over a single pillar accumulates and the sum-total of all pillar interactions translates to a larger sensible change in D.sub.app measured at the output of the DLD devices.

[0119] This cell deformation at increasing fluid flow rates is the leading cause of the decreasing D.sub.app. Detailed analyses of WBC velocity for each pillar structure provide a deeper understanding of cell-DLD interactions (see FIG. 6).

[0120] WBC Biophysical Signatures in DLD Assay

[0121] Unlike rigid beads, the WBC sorting differences for L and L.sup.−1 DLD assays give rise to unique biophysical signatures. It was hypothesized that the unique biophysical signatures can be utilized to profile WBC samples. To investigate the variations of the biophysical signature, the DLD assays were performed at the same flow parameters on 5 healthy donor samples, and the results are shown in FIGS. 7A and 7B. L.sup.−1 D.sub.app measures consistently higher at 9.98±0.15 μm than L D.sub.app at 9.37±0.21 μm at 2.5 μL/min with a p-value of 0.004. The CDM measurements for the 5 healthy samples also showed consistent differences with CDM.sub.L-1 having a larger value of 1.19±0.13 compared to CDM.sub.L of 0.85±0.07 with a p-value of <0.001. Both tests were performed using a paired 2-tailed t-test.

[0122] To evaluate the biophysical combinatorial immune profiling potential of DLD assay, a single biophysical size and deformability parameter is determined. For size parameters, the average D.sub.app for L and L.sup.−1 assays was quantified at 2.5 μL/min, while a single cell deformability parameter (CDM.sub.dot) was emphasized by taking the product of the CDM.sub.L and CDM.sub.L-1 measurements. Performing a product amplifies the deformability differences compared to CDM.sub.L and CDM.sub.L-1 measurements individually (see FIG. 8). By combining the measurements of two DLD devices, a unique leukocyte biophysical fingerprint as a profiling tool to measure different states of WBC from whole blood in real-time was envisioned (see FIG. 1B).

[0123] DLD Assay Combinatorial Immune Profiling

[0124] Using the two parameters of D.sub.app and CDM.sub.dot, various conditions of immune cells from whole blood were profiled. Three groups of immune cell conditions were measured, namely direct sample sparing measurements, in vitro WBC assays and impact of blood processing methods on immune cell biophysical properties (FIG. 7C).

[0125] Direct sample measurements enable the study of WBC biophysical profiles in different health states of an individual for healthy donors, patients admitted to the ED without signs of infection (ED Control), and patients with clear signs of infection and fulfilling at least two SIRS criteria (ED 2 SIRS) (See Table 3). The second group are tests performed on blood, which have undergone external or in vitro test conditions to activate the WBCs. These include 5 ng/mL lipopolysaccharide (LPS), which mimics the bacteria coat that would trigger inflammation and phorbol 12-myristate 13-acetate (PMA) at 100 and 1000 nM, a known activator of WBCs. Lastly, standard blood processing methods were tested, specifically the commonly used RBC lysis protocol to retrieve immune cells and whole blood stored on ice. All tests were initiated within 1 hour of blood draw to accommodate the transport time of blood from hospital to laboratory and concluded within 15 minutes of biophysical profiling.

TABLE-US-00002 TABLE 3 Initial Patient Recruitment for Biophysical Testing Infection Class 0 1 Sub-Class −1 0 1 Characteristic Healthy Control ≥2 SIRS Total Number of patients 5 5 5 10 Number of blood samples 5 5 5 10 Age - mean (years) 34 45 52.2 48.6 Male sex - no. 3 3 2 5 Septic - no. N/A 0 1 1 Septic Shock - no. N/A 0 0 0 Acute renal failure - no. N/A 0 1 1 Death - no. N/A 0 0 0 ≥22 SIRS - no. N/A 0 5 5 Site of infection at the time of blood sampling A. Pneumonia N/A N/A 1 1 B. Urinary tract infection N/A N/A 1 1 C. Intraabdominal infection N/A N/A 2 2 D. Skin or soft-tissue infection N/A N/A 0 0 E. Intrathoracic infection N/A N/A 1 1

[0126] The dotted line in FIG. 7D denotes the baseline mean measurement of D.sub.app for 5 healthy samples at 9.7±0.1 μm. In comparison, for all tested conditions, WBC D.sub.app only seems to increase and not decrease with varying magnitude for different conditions. For the closest in vivo model of direct sample injection, ED≥2 SIRS showed a larger size of 10.4±0.4 μm compared to ED control (9.8±0.3 μm). In contrast, WBC biochemical activation with incubation of PMA at 100 nM and 1000 nM for 2 hours showed a much larger increase of D.sub.app at 12.6±0.8 μm and 13.9±0.7 μm, respectively. This is a drastic 30% to 43% increase in WBC D.sub.app. Interestingly, the RBC lysis process also showed an increase in WBC D.sub.app, which suggests potential activation and biophysical changes in the WBCs. Blood processing protocol to store blood samples on ice did not change the WBC D.sub.app.

[0127] CDM.sub.dot measurements on the other hand describe a different parameter of the WBC. A larger CDM.sub.dot relative to the healthy donor measurements in FIG. 7E indicates an increase in deformability while a reduced CDM.sub.dot shows a decrease in deformability. These CDM differences are benchmarked against the control measurements of healthy samples with a CDM.sub.dot of 0.98. A surprising result emerged from the deformability parameter measurement as all in vitro assays and sample pre-processing steps resulted in reduced deformation of WBC measured. This means that the size compression of WBC is reduced compared to the healthy baseline. On the contrary, direct sample injection tests showed that the WBC deformability increased instead when infection is present in the patient. The greatest change in deformability was demonstrated in samples of blood on ice where the CDM.sub.dot is greatly reduced to less than 0.10 while its size measurements did not change. The cold temperatures experienced by cells could potentially affect mechanical properties of cells due to stiffening of cytoskeletal networks and lipid bilayer.

[0128] WBCs from patients who have infections show an increase in deformation relative to WBCs of healthy donors. The divergence of CDM.sub.dot measurements was unexpected. This suggest that in vitro assays mimicking WBC activation could not replicate the physiological conditions of WBC biophysical parameters despite incubation in whole blood at 37° C., as post-blood draw WBC activation assays illicit a different biophysical response relative to innate blood from infected patients. This evidently shows that activation of WBC is multi-dimensional and complex physiologically. Simple and single triggers of activation are highly unlikely the cause for the observed WBC biophysical characteristics. The data also emphasize the conflicting results of earlier studies showing WBCs of ICU sepsis patients being less deformable while other previous works showed increase in WBC deformability during infection. Previous studies also attempted to mimic sepsis via biochemical trigger cocktails but were unable to do so. This highlights the importance of the various exemplary embodiments disclosed herein in developing tools to probe innate immune states with minimal sample handling and ex vivo delay time.

[0129] Biophysical Immune Markers of Severe Inflammation

[0130] Patient recruitment from the ED of the National University Hospital, Singapore, was conducted with ethics approval from the local institutional review board (National Healthcare Group Singapore, Domain Specific Review Board, DSRB reference number: 2018/00115). Written informed consent was obtained from enrolled participants. Based on the results from FIGS. 7A to 7E, the direct sample injection tests were expanded and 85 donors comprising two broad categories of donors without infection or with infection were recruited. The “no infection” group includes 8 healthy donors (Healthy) and 36 ED admission controls (Controls) while the “infection” group comprised 41 ED admissions with infection plus ≥2 SIRS criteria (see Table 4). Additionally, instead of just size and deformability parameters, cell count and the WBC distribution were evaluated with a range of 38 identified biophysical markers parameters (features) listed from the DLD assay resulting in a clearly distinct PCA plot (See FIG. 9 and Tables 5 and 6).

TABLE-US-00003 TABLE 4 Larger Cohort Test Patient Recruitment from ED. Infection Class 0 1 Sub-Class −1 0 1 2 Characteristic ≥2 SIRS + Healthy Control ≥2 SIRS Sepsis Total Number of patients 8 36 37 4 85 Number of blood 8 36 37 4 85 samples Age - mean (years) 28.7 48.6 54.7 63 50.4 Age - S.D. (years) 3.7 10.5 14.0 8.3 13.9 Male sex - no. 3 17 19 3 42 Sepsis - no. N/A 0 0 4 4 Septic Shock - no. N/A 0 0 0 0 Acute renal failure - no. N/A 0 7 2 9 Death - no. N/A 0 0 0 0 ≥22 SIRS - no. N/A 0 37 4 41 Site of infection at the time of blood sampling A. Pneumonia N/A N/A 8 0 8 B. Urinary tract N/A N/A 5 1 6 infection C. Intraabdominal N/A N/A 8 1 9 infection D. Skin or soft- N/A N/A 12 0 12 tissue infection E. Intrathoracic N/A N/A 3 2 5 infection

TABLE-US-00004 TABLE 5 Showing the statistical comparison of 38 features with cross tested paired 2-tailed t-test statistic. Bold numbers show statistical significance. Feature Statistic Feature p-value No. Name Description Class −1 vs 0 Class 0 vs (1 + 2) Class −1 vs (1 + 2)  1 Size 1 L.sub.−1 D.sub.app @ 2.5 μl/min 0.0997 <0.00001 0.00001  2 Size 2 L.sub.−1 D.sub.app @ 25.0 μl/min 0.9566 0.06115 0.30468  3 Size 3 L D.sub.app @ 2.5 μl/min 0.0111 <0.00001 0.00002  4 Size 4 L D.sub.app @ 25.0 μl/min 0.0224 0.00998 0.00883  5 Size 5 <L.sub.−1 D.sub.app> 0.2288 0.00002 0.00248  6 Size 6 <L D.sub.app> 0.0011 <0.00001 0.00004  7 Size 7 Average D.sub.app @ 2.5 μl/min 0.0124 <0.00001 0.00001  8 Size 8 Average D.sub.app @ 25.0 μl/min 0.1091 0.0140 0.0472  9 Size 9 Stdev of L.sub.−1 D.sub.app @ 2.5 μl/min 0.8531 0.1072 0.4378 10 Size 10 Stdev of L.sub.−1 D.sub.app @ 25.0 μl/min 0.4445 0.0147 0.0450 11 Size 11 Stdev of L D.sub.app @ 2.5 μl/min 0.7798 0.2521 0.4147 12 Size 12 Stdev of L D.sub.app @ 25.0 μl/min 0.2769 0.2758 0.0691 13 Def 1 Size 1-Size 3 0.3325 0.0650 0.1256 14 Def 2 Size 2-Size 4 0.0659 0.3172 0.0110 15 Def 3 (Size 1-Size 3)/(Size 2-Size 4) 0.4121 0.4294 0.7227 16 Def 4 CDM.sub.L−1 0.2073 0.0032 0.0072 17 Def 5 CDM.sub.L 0.9030 0.0015 0.1069 18 Def 6 CDM.sub.L: CDM.sub.L−1 0.3855 0.5156 0.6098 19 Def 7 CDM.sub.dot 0.5441 0.0003 0.0234 20 Def 8 Size 6-Size 5 0.1202 0.0663 0.0049 21 Dist 1 Skew: L.sub.−1 D.sub.app @ 2.5 μl/min 0.0781 0.0874 0.0023 22 Dist 2 Skew: L.sub.−1 D.sub.app @ 25.0 μl/min 0.7113 0.0201 0.3623 23 Dist 3 Skew: L D.sub.app @ 2.5 μl/min 0.7554 0.8006 0.6018 24 Dist 4 Skew: L D.sub.app @ 25.0 μl/min 0.3362 0.1645 0.9630 25 Dist 5 Dist 1-Dist 2 0.5413 0.0025 0.0259 26 Dist 6 Dist 3-Dist 4 0.2847 0.3360 0.6660 27 Dist 7 Dist 1-Dist 3 0.3668 0.3035 0.0692 28 Dist 8 Dist 2-Dist 4 0.3598 0.0003 0.2394 29 Dist 9 Kurtosis: L.sub.−1 D.sub.app @ 2.5 μl/min 0.1565 0.4656 0.0197 30 Dist 10 Kurtosis: L.sub.−1 D.sub.app @ 25.0 μl/min 0.6935 0.0338 0.4194 31 Dist 11 Kurtosis: L D.sub.app @ 2.5 μl/min 0.8214 0.4844 0.8445 32 Dist 12 Kurtosis: L D.sub.app @ 25.0 μl/min 0.2152 0.1328 0.8593 33 Dist 13 Dist 1 * Dist 9 0.8708 0.6188 0.8843 34 Dist 14 Dist 9 * Dist 11 0.3060 0.8689 0.4149 35 Count 1 Cell Count: L.sub.−1 D.sub.app @ 2.5 μl/min 0.6253 <0.00001 0.0065 36 Count 2 Cell Count: L.sub.−1 D.sub.app @ 25.0 μl/min 0.6469 0.00022 0.0246 37 Count 3 Cell Count: L D.sub.app @ 2.5 μl/min 0.9413 0.00002 0.0260 38 Count 4 Cell Count: L D.sub.app @ 25.0 μl/min 0.5729 0.00003 0.0157

TABLE-US-00005 TABLE 6 Description of the features and identified 38 selected markers for profiling of WBC using the DLD assay. Feature Statistic No Feature Name Description Remarks  1 Size 1 L.sub.−1 D.sub.app @ 2.5 μl/min WBC size measured using L.sup.−1 DLD at 2.5 μl/min (Slow)  2 Size 2 L.sub.−1 D.sub.app @ 25.0 μl/min WBC size measured using L.sup.−1 DLD at 25 μl/min (Fast)  3 Size 3 L D.sub.app @ 2.5 μl/min WBC size measured using L DLD at 2.5 μl/min (Slow)  4 Size 4 L D.sub.app @ 25.0 μl/min WBC size measured using L DLD at 25 μl/min (Fast)  5 Size 5 <L.sub.−1 D.sub.app> Mean size measure at 2.5 and 25 μl/min for L.sup.−1 DLD  6 Size 6 <L D.sub.app> Mean size measure at 2.5 and 25 μl/min for L DLD  7 Size 7 Average D.sub.app @ 2.5 μl/min Mean size of WBC at 2.5 μl/min for L and L.sup.−1 DLD  8 Size 8 Average D.sub.app @ 25.0 μl/min Mean size of WBC at 25 μl/min for L and L.sup.−1 DLD  9 Size 9 Stdev of L.sub.−1 D.sub.app @ 2.5 μl/min Standard deviation of WBC size for L.sup.−1 DLD at 2.5 μl/min 10 Size 10 Stdev of L.sub.−1 D.sub.app @ 25.0 μl/min Standard deviation of WBC size for L.sup.−1 DLD at 25 μl/min 11 Size 11 Stdev of L D.sub.app @ 2.5 μl/min Standard deviation of WBC size for L DLD at 2.5 μl/min 12 Size 12 Stdev of L D.sub.app @ 25.0 μl/min Standard deviation of WBC size for L DLD at 25 μl/min 13 Def 1 Size 1-Size 3 Comparing size difference between L and L.sub.−1 14 Def 2 Size 2-Size 4 Comparing deformed difference between L and L.sub.−1 15 Def 3 (Size 1-Size 3)/(Size 2-Size 4) Ratio of L and L.sup.−1 differences of size vs deformed state 16 Def 4 CDM.sub.L−1 Deformation modulus of L.sub.−1 17 Def 5 CDM.sub.L Deformation modulus of L 18 Def 6 CDM.sub.dot Product of deformation modulus 19 Def 7 CDM.sub.L: CDM.sub.L−1 Ratio of deformation modulus 20 Def 8 Size 6-Size 5 Combined size and deformability difference of L and L.sub.−1 21 Dist 1 Skew: L.sub.−1 D.sub.app @ 2.5 μl/min The skew of the histogram for L.sup.−1 at 2.5 μl/min 22 Dist 2 Skew: L.sub.−1 D.sub.app @ 25.0 μl/min The skew of the histogram for L.sup.−1 at 25 μl/min 23 Dist 3 Skew: L D.sub.app @ 2.5 μl/min The skew of the histogram for L at 2.5 μl/min 24 Dist 4 Skew: L D.sub.app @ 25.0 μl/min The skew of the histogram for L at 25 μl/min 25 Dist 5 Dist 1-Dist 2 Skew differences between size and deformed states for L.sub.−1 26 Dist 6 Dist 3-Dist 4 Skew differences between size and deformed states for L 27 Dist 7 Dist 1-Dist 3 Skew differences of size histogram of L.sup.−1 and L 28 Dist 8 Dist 2-Dist 4 Skew differences of deformed state of L.sup.−1 and L 29 Dist 9 Kurtosis: L.sub.−1 D.sub.app @ 2.5 μl/min The Kurtosis of the histogram for L.sup.−1 at 2.5 μl/min 30 Dist 10 Kurtosis: L.sub.−1 D.sub.app @ 25.0 μl/min The Kurtosis of the histogram for L.sup.−1 at 25 μl/min 31 Dist 11 Kurtosis: L D.sub.app @ 2.5 μl/min The Kurtosis of the histogram for L at 2.5 μl/min 32 Dist 12 Kurtosis: L D.sub.app @ 25.0 μl/min The Kurtosis of the histogram for L at 25 μl/min 33 Dist 13 Dist 1 * Dist 9 Product amplification of skew and Kurtosis for L.sub.−1 34 Dist 14 Dist 9 * Dist 11 Kurtosis product of size measured at L.sup.−1 and L 35 Count 1 Cell Count: L.sub.−1 D.sub.app @ 2.5 μl/min Cell count per frame at L.sup.−1 flow 2.5 μl/min 36 Count 2 Cell Count: L.sub.−1 D.sub.app @ 25.0 μl/min Cell count per frame at L.sup.−1 flow 25 μl/min 37 Count 3 Cell Count: L D.sub.app @ 2.5 μl/min Cell count per frame at L flow 2.5 μl/min 38 Count 4 Cell Count: L D.sub.app @ 25.0 μl/min Cell count per frame at L flow 25 μl/min

[0131] By plotting the correlation heat map for all the biomarkers, the various correlation clusters of the biomarkers can be distinguished. A 2-tailed t-test was performed for the results shown in FIGS. 10A to 10H and Table 5. Interestingly, it was found that the L DLD assay in FIGS. 10C and 10D could significantly detect differences between all three inter-sample groups (sub-class) of healthy, control and 2 SIRS. Cell count data (shown in FIGS. 10E to 10H) only showed distinct differences between control and 2 SIRS group. This is not unexpected as cell count is a parameter to assess SIRS criteria. However, it is highly intriguing that cell size (specifically L pillar in Size 3 and Size 4; see FIGS. 10C and 10D) were able to differentiate healthy from ED control group. This suggests that despite no clear signs of infection, the immune size is modulated based on medical conditions that the patients were admitted for. The mean cell size difference between the three groups range from 9 to 10 μm which validates the resolution of DLD assay to probe cells within this narrow D.sub.app band. The corresponding statistical analysis for the deformability and distribution markers also showed significant distinction between no-infection and infection group (see Table 5). The correlation heatmap shows that these markers are independently significant as they are not strongly correlated and can be used for immune profiling (see FIG. 11).

[0132] The 38 biophysical markers of all tested samples were tabulated and hierarchical clustering was performed based on the DLD assay biophysical markers (FIG. 10I). The unsupervised clustering grouped the data into 8 clusters with visible biophysical signatures and profiles. Patients with >2 SIRS were generally clustered in group 1-4 while non-infection control and healthy donors were grouped in cluster 5-8. Visible distinction between these groups can be seen in the heatmap. Size and deformability-based biomarkers were elevated for >2 SIRS group while cell distribution biomarkers levels were cluster specific. Cell count biomarker were only highly expressed in certain samples and is not directly correlated with size-based markers for group 3 and 4. This suggest that biophysical markers such as immune cell size and deformability is potentially more sensitive to profile the immune activation states than lagging changes in cell count.

[0133] Interestingly, 4 patients from >2 SIRS group were later diagnosed to have sepsis with a SOFA score of >4 for >2 SIRS 03, 16, 22 and 35 in FIG. 10I and were grouped in cluster 2. This group showed moderate increase in size and cell count but relatively larger increase in cell deformability and distribution markers. Cluster group 2 also correlated a longer hospitalization stay and this biophysical signature could be useful to prognose severity of disease progression and risk of hospitalisation (See FIG. 12).

[0134] >2 SIRS 02, 08 and 13 immune signatures were clustered in group 5-8 which was predominantly healthy and non-infection controls. This clustering independently shows that these patients, though exhibiting >2 SIRS, had a lower immune response signature profile which resulted in a short hospitalisation stay of only 1-2 days. On the contrary, Control sample 19 in cluster 1 had a relatively longer hospitalisation stay of 10 days. Finally, the predictive value of 38 biophysical markers to classify non-infection versus infection class of 85 patient samples was analysed using the receiver operating characteristic (ROC) curve in FIGS. 13 and 14. The area under curve (AUC), specificity and sensitivity of the assay was 0.97, 0.91 and 0.92 respectively. This result was performed based on support vector machine model for classification (FIG. 14)

[0135] The results discussed above show that the DLD devices function as sensitive and quantitative assay of immune cell biophysical signatures in relation to the WBCs' real-time activation levels. The swift response of the immune system induced by biochemical triggers are also expressed in biophysical properties of the leukocytes for effective extravasation and other functions. Studies have shown correlation of immune cell biophysical changes with cytoskeletal remodelling, protein production and cell proliferation. As WBCs are activated by various internal or external triggers, the extent and direction of these changes were sensitively measured using the DLD assay described in various exemplary embodiments. The results highlight new insights, which advances both engineering of precision microfluidics and clinical research.

[0136] Applications

[0137] First, various DLD structures were shown to illicit different sorting signatures on deformable cells. The selection of L and L.sup.−1 was not arbitrary as it is based on previous observations on RBC sorting performance. Various embodiments of the present disclosure can entail the possibility that more suitable DLD pillar shapes can exist for the function of biophysical DLD assays. To uncover potentially useful DLD shapes requires deeper and fundamental understanding of particle-pillar-fluid interactions, especially for deformable particles. The empirical evidence and simulations discussed show that using these different signatures, a collective cell D.sub.app and deformability response that quantitatively predicts a cell state can be defined.

[0138] Second, the WBC biophysical DLD assay showed divergent deformability response for in vitro assays and direct whole blood assay. In vitro assays here, which aim to study WBC immune response, were not able to replicate the biophysical deformability properties of WBC from patients who show clear signs of infection. This could be due to blood treatment methods using ethylenediaminetetraacetic acid (EDTA), stimulants concentration and incubation time. Recent advances in microfluidic devices based on high-throughput single cell deformability imaging cytometry mechano-phenotyping also showed that natively activated immune cells increases its deformability and size and also showed oscillating immune activity during immune activation and sepsis. Similarly, the results discussed based on whole blood rapid immune profiling supports this crucial finding and raises new research questions and potentially challenging current methods of using in vitro studies to elucidate physiological immune responses.

[0139] Finally, the clinical study discussed shows that patient classification using DLD biophysical assay was possible showing distinct label-free biomarker profiles of healthy donors and patients admitted to the emergency department with and without infection. The approach adopted differs substantially from previous ICU-based studies where patients who have clear manifestations of symptoms and signs of severe disease and immune dysregulation. Importantly, these patients were recruited at admission to the ED with diverse pre-existing conditions such as diabetes mellitus and hypertension but did not progress to full-blown sepsis, characterized by presence of organ dysfunction. Yet, immune biophysical markers show independent and good indication of its diagnostic or prognostic potential, especially the possibility for identifying patients with non-infection-related medical conditions. While a larger clinical study is needed to further evaluate potential biophysical immune response phenotypes and its utility in the field, the study discussed adds scientific evidence to existing works on biophysical parameters as an important marker for immune profiling.

[0140] Various embodiments of the present disclosure provide unique biophysical signatures when immune cells are sorted from whole blood within unconventional DLD pillars of L and L.sup.−1 shape. These signatures result in the formulation of 38 biophysical markers which enable the profiling of immune responses of patients recruited from emergency department with a detection sensitivity of 0.91 and specificity of 0.92. Given that the DLD assay in various embodiments disclosed herein takes 15 minutes to perform, uses less than 20 μL of whole blood and only requires video capture frame rates of up to 150 fps, the system can potentially be developed into a portable unit for point-of-care whole blood sparing assays which could significantly improve the diagnosis and stratification of patients with systemic inflammation response syndrome within the ED and other primary care settings. The availability of such an adjunct with both real-time information and rapid turnaround time is crucial as incoming patients to the ED from the community are highly undifferentiated. Being able to quickly identify at-risk patients and render measures to prevent organ dysfunction will be the key actionable information provided by this tool. This contrasts with patients in the ICU who already have clinical evidence of organ dysfunction through standard laboratory investigations and physiological parameters.

Experimental Section/Methods

[0141] The methods taken for conducting the DLD assays in accordance with various embodiments disclosed herein are provided as follows.

[0142] Device Design

[0143] DLD is a sensitive size-based sorting technique, using a regularly spaced pillar array where the separation can be determined by the established empirical formula:

D.sub.c=1.4 G tan θ.sup.0.48  (1)

Where G is the regular spacing between pillars and θ is the offsetting angle of the pillars. Two DLD chips with 21 DLD segments to compare L and L.sup.−1 shape DLD pillars were designed. The G used measures 23 μm and with D.sub.c of device ranging from 6.0 to 16.0 μm, each DLD segment increases the D.sub.c by a step of 0.5 μm. The period of the array is 50 μm.

[0144] Device Fabrication

[0145] The device was fabricated using standard photolithography methods. A chromed quartz mask with the designs specified was ordered from JD Photo Data (Hitchin, UK). A mask aligner was used to fabricate an SU-8 mold using SU-8 2015 and spun to a thickness of approximately 20 μm. Poly-dimethylsiloxane (PDMS) (Dow Corning, Midland, Mich.) was added in a ratio of 1:10 and poured onto the SU-8 master mold. The PDMS was cured into an oven at 75° C. for 1 hour to crosslink the PDMS. Finally, the PDMS was peeled out of the master mold and cut into the dimensions of the DLD chip.

[0146] Three 3 mm holes were punched as inlet reservoirs to hold the blood sample and 1×PBS buffer. A 1.5 mm punch was used in the outlet to connect the device to the tubing and syringe. Finally, the device was bonded onto a glass slide using oxygen plasma surface activation and bonding. The chip was ready to be used the next day.

[0147] An exemplary system for conducting DLD assays is shown in FIG. 15. The system 1500 comprises a DLD device 1502 (compare DLD device 100 of FIG. 1A). The DLD device 1502 comprises an inlet port to a sample reservoir 1504A (compare sample/open reservoir 102A of FIG. 1A) and inlet ports to buffer reservoirs 1504B and 1504C (compare buffer/open reservoirs 102B and 102C of FIG. 1A). The DLD device 1502 further comprises an outlet port 1506 (compare outlet port 104 of FIG. 1A). In the exemplary embodiment, the DLD device 1502 is mounted on a detection set up in the form of a camera, lens and detector housing 1508 and has a light source 1510 in the vicinity. The DLD device 1502 is further coupled to a waste collector 1512 via the outlet port 1506 for collecting waste. The waste collector 1512 is further coupled to a filter 1514, control valves 1516 and a syringe/pressure pump 1518. In the exemplary embodiment, the syringe/pressure pump 1518 is configured to control or regulate the flow rates of fluids flowing through the reservoirs of the DLD device 1502. The system 1500 further comprises a switch and power source 1520, function buttons 1522 and a pressure/flow reader and screen 1524. In the exemplary embodiment, the power source 1520 and the function buttons 1522 are configured to control the syringe/pressure pump 1518, i.e., to control the flow rates of fluids flowing through the reservoirs of the DLD device 1502. The pressure/flow reader and screen 1524 is configured to display measured pressure and/or flow readings.

[0148] Reagents

[0149] The beads used were size calibration standards kit 6.2, 7.2, 8.3 and 10.2 μm beads from Bangslab (Bangs Laboratories, Fishers, Ind.). They were resuspended (2 million mL.sup.−1) to 25 be used in the characterisation tests. Lipopolysaccharides from Escherichia coli 0111:64 (L2630) and Phorbol 12-myristate 13-acetate (P8139) were purchased from Merck-Sigma (St Louis, Mo.). The LPS concentration (5 ng/mL) was determined based on previous works. 1× phosphate buffer solutions were used for all dilutions of beads and as sample buffer.

[0150] Donor Selection Criteria

[0151] Patient recruitment from the ED of the National University Hospital, Singapore, was conducted with ethics approval from the local institutional review board (National Healthcare Group Singapore, Domain Specific Review Board, DSRB reference number: 2018/00115). Written informed consent was obtained from enrolled participants.

[0152] The ED controls in the study comprised of patients who attended the ED for symptoms 10 unrelated to inflammatory or infectious conditions such as corneal foreign body, poorly controlled hypertension while the healthy volunteers included fellow colleagues working in the ED. These two groups of donors constitute the “no infection” group (Infection Class=0).

[0153] For the “infection” group (Infection Class=1), patients who had a clear and objective source of systemic infection based on preliminary investigations such as chest radiography, urine or blood investigations and fulfils at least 2 SIRS criteria (fever >38 or <36 degrees Celsius; respiratory rate >20/min; heart rate >90/min; white blood cell count >12,000/mm3, <4,000/mm3, or >10% bands) were enrolled.

[0154] Vulnerable population (such as pregnant or incarcerated individuals), patients less than 21 years old, those who refused or were unable to provide written informed consent and patients with “do-not-resuscitate” orders were excluded. Additionally, patients with medical conditions or medications that may result in macrocytosis were also excluded as this could potentially interfere with evaluation of WBC size and deformability. These include conditions such as vitamin B12 deficiency, primary bone marrow disorder, previous gastrectomy, pernicious anemia, alcoholism, COPD, familial macrocytosis, hypothyroidism, cancer and medications like chemotherapy agents, zidovudine, trimethoprim, phenytoin and oral contraceptive pills.

[0155] Blood Collection and Testing

[0156] All blood collected were from venous blood draw with consent from patients at the ED of National University Hospital, Singapore. Post-recruitment, the blood (3 mL) was drawn into a 3 mL EDTA tube and stored in a cooler box to maintain the temperature. The transport of blood from draw to laboratory experiments was within 1 hour. Blood samples (100 μL) was aliquoted out for each test.

[0157] Activation of Leukocytes

[0158] All WBC experiments, if not tested immediately, were placed on 37° C. water bath to ensure physiological conditions. There were no dilutions of blood. LPS activation test (5 ng/mL) was incubated for 30 minutes. As each test run was 15 minutes, more vials were prepared in time spacing of 3 minutes each for testing of each flow rate. This was to ensure the tests were performed at 30 minutes interval and the data acquisition time was not a factor. For PMA activation (100 nM and 1000 nM), the samples were incubated for 2 hours. All sample predilutions were made on 1×PBS.

[0159] Data Acquisition and Analysis

[0160] A Phantom V7.1 (Vision Research, Wayne, N.J.) was used to capture all visual data from input, output and single cell motion within all DLD devices. The video files were exported into uncompressed “.avi” format for downstream analyses and counting. For each experiment, a total of 2500 frames were captured for analysis. The frame rates used for capture were 15, 30, 60 and 150 fps for 2.5, 5.0, 10.0 and 25.0 μL/min flow rates, respectively. The analysis of cell 20 counting to plot the histogram was performed by a custom python code, which plots the counted cells against the sub-channel location. From the normalized frequency distribution histogram, the mean, S.D., skew, Kurtosis, frequency, and distribution data were available.

[0161] Machine Classification

[0162] Hierarchical clustering and PCA analysis were all performed using python 3.6 with module “scikit-learn”. To develop the ROC curve, a custom algorithm shown in FIG. 14 coupled with support vector machine (SVM) classification using radial basis function kernel was used.

[0163] The algorithm used to calculate diagnostic probability values of each sample are shown in FIG. 14. Each sample is selected as a “blind” sample and the remaining (n=84) samples are randomly split 9:1 part for boot strapping method of 1000 cycles to validate the prediction of the “blind” sample based on SVM classification. The boot strapping method results in a probability value of predicting the class of “blind” sample. The probability would then be fed into the ROC curve and comparing with its known class for sensitivity and specificity calculation.

[0164] Flow Coupled Cell Simulations

[0165] Deformable 2D cell simulations were carried out with the help of a bespoke lattice-Boltzmann-immersed-boundary code. The algorithm is well established for particulate flows in the low Reynolds number regime. The 2D cell is modelled as a ring of marker points that deform according to well defined physical energy potentials.

[0166] It will be appreciated by a person skilled in the art that other variations and/or modifications may be made to the embodiments disclosed herein without departing from the spirit or scope of the disclosure as broadly described. For example, in the description herein, features of different exemplary embodiments may be mixed, combined, interchanged, incorporated, adopted, modified, included etc. or the like across different exemplary embodiments. The present embodiments are, therefore, to be considered in all respects to be illustrative and not restrictive.