Analytical validation of the DropXpert S6 system for diagnosis of chronic myelocytic leukemia

Abstract

Digital PCR is a powerful method for absolute nucleic acid quantification and is widely used in the absolute quantification of viral copy numbers, tumor marker detection, and prenatal diagnosis. However, for most of the existing droplet-based dPCR systems, the droplet generation, PCR reaction, and droplet detection are performed separately using different instruments. Making digital PCR both easy to use and practical by integrating the qPCR workflow into a superior all-in-one walkaway solution is one of the core ideas. A new innovative and integrated digital droplet PCR platform was developed that utilizes cutting-edge microfluidics to integrate dPCR workflows onto a single consumable chip. This makes previously complex workflows fast and simple; the whole process of droplet generation, PCR amplification, and droplet detection is completed on one chip, which meets the clinical requirement of “sample in, result out”. It provides high multiplexing capabilities and strong sensitivity while all measurements were within the 95% confidence interval. This study is the first validation of the DropXpert S6 system and focuses primarily on verifying its reliability, repeatability, and consistency. In addition, the accuracy, detection limit, linearity, and precision of the system were evaluated after sample collection. Among them, the accuracy assessment by calculating the absolute bias of each target gene yielded a range from −0.1 to 0.08, all within ±0.5 logarithmic orders of magnitude; the LOB for the assay was set at 0, and the LoD value calculated using probit curves is MR4.7 (0.002%); the linearity evaluation showed that the R² value of the BCR-ABL was 0.9996, and the R² value of the ABL metrics calculated using the ERM standard was 0.9999; and the precision evaluation showed that all samples had a CV of less than 4% for intra-day, inter-day, and inter-instrument variation. The CV of inter-batch variation was less than 7%. The total CV was less than 5%. The results of the study demonstrate that dd-PCR can be applied to molecular detection and the clinical evaluation of CML patients and provide more precise personal treatment guidance, and its reproducibility predicts the future development of a wide range of clinical applications.

---

Introduction

Digital PCR is a powerful method for absolute nucleic acid quantification with unprecedented accuracy and precision and is widely used in absolute quantification of viral copy numbers, copy number variant detection, tumor marker detection, and prenatal diagnosis.^1–4 Currently, available commercial digital PCR platforms rely on two distinct approaches. The first to be developed was well-based digital PCR (dPCR), which relies on 2D arrays of microwells to partition the sample.² Once filled with the PCR mix, the microchambers are thermocycled on a flat-block thermocycler and then imaged using fluorescence to identify the amplified positive partitions. The typical commercialized digital droplet PCR platforms are the microfluidic-chamber-based QuantStudio AbsoluteQ from Thermo Fisher and Digital Lightcycler from Roche. The second approach is droplet digital PCR (ddPCRTM), in which the sample is partitioned in a bulk emulsion of microdroplets using a specific microfluidic chip.^5,6 The emulsion is transferred to a PCR tube or distributed in one layer in a microchamber for thermocycling. After PCR amplification, data acquisition is performed in a process analogous to flow cytometry, whereby droplets are fluorescently read one by one or imaged using a camera.³ There are two main digital droplet PCR platforms: the droplet-based ddPCR (ddPCR) QX200 from Bio-Rad® and Crystal Digital PCR from Stilla.^4,7 Over these platforms, the droplet-based dPCR has main advantages over other methods. First, the droplet-based dPCR usually has approximately 20 000 to 10 000 000 partitioned droplets per reaction which has higher sensitivity and accuracy than the other two dPCR platforms.^8–10 Secondly, the droplet-based dPCR can realize high throughput and low price. However, for most of the existing droplet-based dPCR systems, the droplet generation, PCR reaction, and droplet detection are performed separately using different instruments. Making digital PCR both easy to use and practical, reducing the time, and improving repeatability by integrating the qPCR workflow into a superior all-in-one walkaway solution, is one of the core ideas.^11–13

In this paper, the invention of the DropXpert S6 digital droplet PCR system has remedied some drawbacks of the previous methods, such as avoiding droplet transfer from one vessel to another or microchip transfer from one machine to another, reducing the high cost of chips and simplifying and automating the whole process. There are 4 main advantages of this novel system: first, it uses syringes instead of an air compressor, to produce the pressure to generate droplets, which makes the equipment more space-friendly and compact. Secondly, by introducing low-density and high boiling point oil as the droplet generation oil, the sample or PCR reaction solution was covered underneath the oil, and sample loss was greatly reduced by making all solution into droplets without introducing air bubbles into chambers, as well as the droplets were covered underneath the low-density oil, preventing droplet evaporation and cross contamination during thermocycling. Thirdly, no droplet transformation is needed which reduces sample loss and makes an incredibly high level of precision. Fourthly, the system is designed to simplify the ddPCR workflow in which only 3 simple steps are needed to have the result you want in less than 2 hours: pipette the sample and load the chip, let the experiment run and automatically analyze data, which makes ddPCR as easy as qPCR.^8,14

In this study, the aim was to validate the analytical performance of the DropXpert S6 system for the detection of the BCR-ABL fusion gene/BCR gene for the diagnosis of chronic myelogenous leukemia using clinical samples. Chronic myeloid leukemia (CML) is a malignant myeloproliferative neoplasm occurring in pluripotent hematopoietic stem cells.⁹ In essence, the C-ABL proto-oncogene on chromosome 9 and the BCR gene on chromosome 22 translocate to each other and form the BCR-ABL fusion gene at the molecular level.^9,10 The p210 protein of the BCR-ABL fusion gene has overwhelmingly strong tyrosine kinase activity,¹⁵ which is encoded by a number of different transcripts. The most common of these are e13a2 (also notated b2a2) and e14a2 (also notated b3a2) accounting for greater than 95% of the CML population. The importance of minimal residual disease (MRD) monitoring in the treatment of hematological malignancies has been unanimously recognized and valued by the hematology community. MRD detection requires the use of highly sensitive and reliable detection methods to improve the detection limit and system stability as much as possible. Its expression level can also be used for the detection of MRD, which has exceedingly important prognostic significance.^15–17 The emergence of tyrosine kinase inhibitors (TKI) targeting BCR-ABL fusion protein has brought the treatment of patients with CML into a new era, significantly improving the prognosis of patients and prolonging the survival time of patients.^18–20 Accordingly, the goal of long-term treatment of CML patients is to discontinue the drug and achieve a treatment-free response (TFR).^21–25 As a consequence, to meet the National Comprehensive Cancer Network (NCCN) guidelines for discontinuing TKI treatment,^26,27 more accurate and sensitive molecular monitoring of the BCR-ABL transcript level has received considerable critical attention for CML patients, so that doctors can better manage patients and make more reasonable treatment plans.^28–31

Experimental and materials

Design and operation of the DropOne C4 chip

As schematically shown in Fig. 1A, the DropOne C4 chip is an injection-molded plastic microfluidic chip that incorporates 4 identical microfluidic networks, each capable of processing one sample or PCR mix. The C4 chip consists of two layers: a top layer containing liquid reservoirs on the upper surface and microchannels and chambers on the lower surface, and a bottom sealing layer using a PC membrane with 250 μm thickness. The two layers of the chip were bonded together through hot-pressing. The main feature of the microfluidic networks is shown in Fig. 1B. Each network has an inlet port, an array of tens of step-emulsification (SE) nozzles, a rectangle chamber, a fence structure, and an outlet. Droplets are generated at the SE nozzles by step emulsification. The end of the SE nozzle has a trapezoid shape with its width gradually increasing from 80 μm to 160 μm. The PCR sample flows out from the SE nozzles and droplet detachment occurs as a result of a local loss of equilibrium due to interfacial tension. Droplets are then stored in the chamber (15 × 21 mm²) for PCR thermocycling. The fence structure prevents droplets from flowing out to the outlet while oil drains out through the fence to the outlet. The C4 chip is designed to produce droplets with a diameter of 100 μm, which equates to a mean droplet volume of 0.5 nL. The C4 chips are prefilled with an emulsion silicone oil containing droplet stabilizing surfactants synthesized by Shenzhen Biorain Biotech Company, such that the entire microfluidic networks and part of the ports are being filled with low-density silicone oil which can cover water droplets underneath to prevent water evaporation during thermocycling. The inlet and outlet ports of C4 chips are sealed with removable Luer caps (Fig. 1C). To prepare the C4 chips, the caps are removed from the inlet ports and 20 μL of PCR mix is pipetted into each of the 4 inlet ports. Pressure-permeable connecting caps (Fig. 1D) are then positioned onto the inlet ports. The pressure-permeable connecting caps are designed with a hole in the middle of caps using PP material and a thin circle on top of the caps using TPE material which can seal between the inlet port of the C4 chip and syringes in the droplet generation unit within the DropXpert S6 instrument.

Fig. 1 The construction of the microfluidic chip DropOne C4. (A) The appearance of the microfluidic chip C4. (B) The feature of the C4 chip with four identical networks. (C) The C4 chip is sealed with Luer caps. (D) The appearance of the gas-permeable connecting caps.

Design and operation of the DropXpert S6 system

The DropXpert S6 system herein is an automated and integrated instrument that can integrate droplet generation, PCR thermocycling, and fluorescence imaging all in one instrument. It was designed and manufactured by Shenzhen Biorain Biotechnology Company including 3 main function units, a droplet generation unit, a thermocycling unit, and a fluorescence imaging unit as illustrated in Fig. 2. Firstly, four DropOne C4 chips were placed on the chip holder of the X–Y moving stage, and the samples were partitioned simultaneously into 2D monolayer arrays of monodisperse droplets using the C4 chips under the pressure of a set of parallel syringes in the droplet generation unit (Fig. 2A). Secondly, after droplet generation, the C4 chips along with the 2-dimension moving stage were moved to the PCR heating module (Fig. 2C) to perform the thermal cycling of the droplet PCR. Lastly, the C4 chips were moved to the fluorescence imaging unit equipped with 4-distinct fluorescence channels (Fig. 2C). The droplet generation unit consists of 4 parallel multi-channel syringes and pressure-down assembly as shown in Fig. 2B which offers enhanced flow performance with high accuracy and smooth flow from 0.1 μL min⁻¹ to 10 μL min⁻¹. The samples are evenly partitioned into tens of thousands of monodisperse droplets under the pressure produced by multi-channel syringes. The thermocline unit consists of 4 independent TEC assemblies, a cooling assembly, and a microchip compressing assembly. After droplet generation, the C4 chips were moved to the thermocline unit by the X–Y moving stage and the chips were pressured up and down by the compressing assembly.

Fig. 2 The DropXpert S6's exterior and interior structure has three main functional units. (A) DropXpert S6 ddPCR instrument appearance. (B) Droplet generation unit: the internal structure of the droplet generation unit includes three main functional units: droplet generation unit, thermal circulator heating unit, and fluorescence imaging unit. (C) Thermocline unit: with four independent temperature control modules. (D) A four-color fluorescence imaging unit using wide-field fluorescence imaging.

Wide-field fluorescence imaging

The four-color fluorescence imaging unit as illustrated in Fig. 3A consists of 1 lens, 1 camera, 3 LED illuminants, 7 filters, 7 dichroic mirrors, 1 focusing motor, and 1 filter switching motor (4-color fluorescence). To view the 15 × 21 mm² area in a single snapshot, a combination of factors was taken into consideration including droplet size, imaging area, and imaging resolution. Ultimately, wide-field fluorescence images are captured using a low-cost 3.1 gigapixel digital camera and macro lens with a 315-degree field-of-view at 1× magnification, respectively. The demonstrated endpoint imaging capability enables on-chip quantitative digital PCR analysis of the entire 15 × 21 mm² droplet chamber. To evaluate the fluorescence uniformity, forty different detection spots as shown in Fig. 3B were selected in one whole droplet storage 0 chamber of the C4 chip, and the uniformity of the fluorescence intensity was measured and exhibited an intensity distribution with a CV of 0.05 (Fig. 3B). The instrument acquires a set of 5 high-resolution images for each chamber containing about 25 000 droplets through the transparent bottom surface of the C4 chips, with one bright-field image and one image for each four-fluorescence acquisition channel (FAM, HEX, ROX CY5). Exposure times for each channel can be changed by the user depending on assay requirements. An image analysis algorithm then identifies each droplet through the bright-field image for its size, position, and shape and characterizes the fluorescence intensity of each identified droplet in each channel through fluorescence images. Artifacts such as dust particles and droplets of inappropriate size are filtered out using size, shape, texture, and position criteria to keep only the fluorescence values of non-coalesced droplets for data analysis.

Fig. 3 Wide field fluorescence imaging features. (A) Representation of the four-color fluorescence imaging unit. (B) In the whole droplet storage chamber of the C4 chip, 40 different detection points were selected to measure the uniformity of fluorescence intensity, and the intensity distribution CV was 0.05. (C) Schematic diagram of the optical structure.

ddPCR workflow

The workflow of ddPCR experiments conducted using the DropXpert S6 system is shown in Fig. 4. After sample preparation, since the C4 chips are already prefilled with oil, you have only one last step before launching the PCR reaction. Briefly, gently pipette 20 μL of PCR mix containing your samples into each inlet port and sealed with pressure-permeable connection caps. The hands-on part is done and ready to place your C4 chips into the DropXpert S6 where the droplet generation, thermal cycling, fluorescence imaging acquisition, and data analysis will be performed sequentially and automatically. Each run can perform 16 samples at once by using 4 chips, and the whole processing will take about 2 hours. The data were analyzed using S6 software which provides intuitive visuals for image analysis, allowing thorough exploration of droplets for quality control. S6 software measures the concentrations of targeted nucleic acids, providing automatic identification of positive and negative droplets for all fluorescence channels. Along with the ability to read data arranged along 1D, 2D, and 3D dot plots, the DropXpert S6 software will give access to raw data to export all experiment details and results.

Fig. 4 DropXpert S6 Digital PCR workflow: after sample preparation, the sample is added and sealed with a connecting cap; the C4 chip is then placed in DropXpert S6, where droplet generation, thermal cycling, fluorescence imaging acquisition, and data analysis are automatically sequentially performed, culminating in data analysis using software.

dPCR reagent preparation

BCR-ABL was used in the study, and ABL was used as the internal reference gene. The materials including PCR primers and probes were designed to reverse transcription and amplify BCR-ABL gene fragments and ABL fragments (Table 1). BCR-ABL's DNA MIX: reverse transcription and cDNA amplification. The reverse transcription reaction consisted of 500 nM of RT primers, 10× RT buffer (Shenzhen Biorain Technology Co., Ltd), and 10 μL of template RNA. The temperature conditions are as follows: 10 min at 25 °C, then 15 reverse transcription at 42 °C followed by 15 s at 85 °C. The polymerase chain reaction consisted of a 20 μL solution containing 500 nM of BCR-ABL primers, 500 nM of ABL primers, 400 nM of specific probes for BCR-ABL, 600 nM of specific probes for ABL, 5× Aplus ddPCR buffer (Shenzhen Biorain Technology Co., Ltd) and 10 μL of template DNA. The PCR reaction mixture is pipetted into C4 chips which are then sealed using pressure-permeable connection caps. The chips are then placed onto the chip holder of the droplet digital PCR system (DropXpert S6). After loading, the program will run the whole process. The samples are partitioned to form thousands of droplets and spontaneously arranged into 2D monolayer arrays of monodisperse droplets at a speed of 2 μL min⁻¹. When thermocycling is completed, the chips are transferred, and images are acquired for the FAM and HEX detection channels. The cycling conditions are as follows: 5 min at 95 °C, then 40 cycles of 10-sec denaturation at 95 °C followed by 30 s extension at 60 °C.

Table 1 Primer information of the experimental materials

Primer & probe	Sequence	Target
RT-primer	ACTCCAAGG	Reverse transcription primer
BCR-ABL forward	TCCGCTGACCATCAATAAGGA	BCR-ABL
BCR-ABL reverse	CACTCAGACCCTGAGGCTCAA
BCR-ABL probe	FAM-CCAGTAGCATCTGAC-MGB
ABL forward	TGTGGCCAGTGGAGATAACA	ABL
ABL reverse	TGGCGTGATGTAGTTGCTTG
ABL probe	VIC-CCAAGGCTGGGTCC-MGB

---

Evaluation of the accuracy

The quantitative accuracy was evaluated using the certified reference material (CRM) ERM-AD623[A]. The measurement for each reference product (n ≥ 2) was repeated, the absolute bias (Bi) according to Bi = Mi − T was calculated by taking the test result (Mi) and logarithmic value of the labeled value of the reference substance (T), and the absolute deviation should not exceed ±0.5 logarithmic orders of magnitude.

Evaluation of the limit of blank (LoB) and the limit of detection (LoD)

The test was repeated twice with 15 CML-negative human blood samples and 15 nucleic acid-free water as blank controls. A non-parametric approach was used to calculate the LOB. The detection rate of BCR-ABL was set by us to be greater than or equal to the 95% limit of detection (LoD). Digital PCR quantification was performed on plasmid samples ZL-E14A2, ZL-E13A2 and ZL-ABL. Three concentration samples were prepared at ratios of 0.005%, 0.002% and 0.001%. Thirty-two replicates were performed for each concentration. The LoD was calculated by using the probit function.

Evaluation of the linearity

Five concentrations of the standard substance ERM-AD623 were used for the linear study, and repeated measurements were made (n = 2). In order to investigate the effect of the transcript type of BCR-ABL 210 on the dPCR system, five concentrations of plasmid samples e13a2 and e14a2, whose transcript types are E13A2 and E14A2 transcripts, were studied linearly, and repeated measurements were performed (n = 3). The plasmid samples were all from Shanghai Bioligo Biotechnology Co., Ltd.

Evaluation of the precision

A total of three concentrations of plasmid samples were used for intra-day, inter-day, inter-instrument, and inter-lot comparisons of assay results. All data were taken to calculate the coefficient of variation. The precision of the reaction system was evaluated by this method.

Evaluation of the clinical sample

A total of 47 patients were included in the present study, and these were diagnosed in the Department of Hematology of Xiangya Hospital of Central South University. The median age of patients was 48.5 years (age range, 11–69 years), including 24 men and 23 women. Clinical samples were collected from peripheral blood from December 25, 2020 to January 17, 2022, and covered all risk levels for both the Sokal score and ELTS score (only 13 patients had Sokal score and ELTS score). Clinical samples of 5 patients from different treatment periods were included. The study scheme was approved by the Ethics Committee of Xiangya Hospital, Central South University, and all patients included in the study provided written informed consent, in accordance with relevant Chinese regulations and the Declaration of Helsinki. The quantitative consistency of the digital PCR and RT-qPCR methods in Bland–Altman plots was evaluated.

Sample processing preparation and nucleic acid extraction

Blood samples were centrifuged at 4000g for 10 minutes to collect the leukocyte layer. BCR-ABL mRNA was extracted using commercially available RNA extraction reagents. TRNzol Universal is produced by Tiangen Biotech (Beijing) Co., Ltd.

Statistical analysis

The results of BCR-ABL transcripts were statistically analyzed, with descriptions of the Bi, Mi, LOB, LOD, mean, and SD. Linear regression analysis was used to analyze the results of ddPCR, and R² represents the coefficient of determination.

Results and discussion

Droplet generation characterization

Through the use of the C4 chip, it is possible to achieve uniform droplet generation at a flow rate of 1–2 μL min⁻¹, producing approximately 25 000 droplets within 20 minutes. The key to accurate ddPCR analysis lies in optimizing the ratio of the number of positive droplets to the total number of droplets. A greater and more stable total number of droplets results in higher accuracy and precision for Poisson-based counting. To assess the stability and repeatability of the droplet generation oil manufactured by Shenzhen Biorain Tech, the sample was amplified (16 amplifications) using 4 chips on the DropXpert S6 platform, targeting the β-actin gene as our amplification target. The total number of droplets and the number of positive droplets for 16 chips were recorded and the corresponding sample concentration was calculated. The results indicated a coefficient of variation (CV) of 4.24% for the total droplet count and 4.48% for the concentration (Table 2). Fig. 5A and C showcase bright-field images of the droplets captured by the DropXpert S6 system, from which 1000 droplets were randomly selected for the study for size analysis by Niko microscopy. The results showed that the average droplet radius before PCR amplification was 59.6 ± 0.3 μm, and that after PCR amplification was 57.7 ± 0.3 μm (Fig. 5B and D). The results showed that the drops were uniform before and after PCR. The coefficient of variation (CV) for the droplet radius was calculated to be 0.74% before amplification and 0.97% after amplification. Overall, the C4 chips with prefilled oil consistently generated approximately 25 000 analyzable droplets, showcasing an impressive theoretical dynamic range spanning 5 orders of magnitude, from 0.2 copies per μL to 20 000 copies per μL, which further validates their exceptional reliability and repeatability.

Table 2 Calculation of the coefficient of variation (CV) for droplet count and concentration in 16 PCR reactions. The CV values provide insights into the variability and precision of the experimental data, allowing us to assess the stability and reliability of the results

Total number of droplets	Number of positive droplets	Concentration (copies per μL)	CV of the total number of droplets	CV of the total number of droplets
21 301	4784	1459.95	4.24%	4.48%
20 108	4842	1581.21
20 351	4849	1562.10
22 475	4984	1439.02
22 123	4975	1462.09
20 473	4749	1514.79
20 315	4688	1505.84
23 227	4831	1338.38
20 930	4884	1525.16
20 410	4967	1600.60
20 565	4597	1452.15
21 533	4980	1509.63
21 806	4931	1471.36
20 683	4972	1578.11
20 451	4913	1576.92
22 345	4968	1659.41

---

Fig. 5 Representation of the droplet states before and after amplification. (A) The droplet state before PCR amplification; (B) statistical analysis indicating the droplet radius before amplification; (C) the droplet state after PCR amplification; (D) statistical analysis representing the expanded droplet radius.

Thermal cycling characterization

Thermal cycling was carried out directly on the droplet by heating from the bottom of the chip. The temperature change was recorded during dPCR amplification as shown in Fig. 6A. The max up-ramp rate of the heating controller is 6 °C s⁻¹ and the down-ramp rate is 7 °C s⁻¹. To investigate the temperature uniformity, the temperature was measured using surfactants on the different spots of the chip during heating, (Fig. 6B). Temperature accuracy was determined by taking the average of these 12 data points. Each accuracy measurement was taken after the instrument had been holding the temperature for 180 seconds. This procedure was repeated 3 times for each of the following temperature set points: 60 and 95 °C. From the temperature distribution histogram, the measured average temperatures on the chip surfaces are 94.9 °C and 59.9 °C at given temperatures of 95 °C and 60 °C, with a temperature accuracy SD of ±0.3 and ±0.3 respectively at different spots of the chip. The temperature distribution is uniform and clearly shows an excellent consistency between measured temperatures and the temperatures displayed on the instrument interface.

Fig. 6 Temperature changes during dPCR amplification and temperature measurement. (A) The heater controls a maximum rise rate of 6 °C s⁻¹ and a maximum fall rate of 7 °C s⁻¹, temperature set-points: 60 °C and 95 °C. (B) The temperature accuracy of the surfactant measurement at the 12 data points selected at different locations of the chip during the heating process. (C) The measured temperature is very consistent with the temperature displayed on the instrument interface. (D) The temperature distribution map, with 144 points at 95 degrees 12 × 3 × 4 and 144 points at 60 degrees.

Visualization of four-color experiments

A key feature of the DropXpert S6 system is the ability to access raw data, the bright-field and fluorescence images of the droplets, which allows the user to perform a preliminary qualitative assessment of the experiment, and check that the stability and the acquisition parameters of droplets, such as droplet uniformity, impurity debris, and exposure times, (Fig. 7). Moreover, advanced features within the S6 analysis software also allow users to visualize which objects have been falsely classified as droplets, or positive events and manually correct it. Another powerful function of S6 software is that it allows users to select wells to merge the positive and negative droplets from all the wells and analyze these data together as a single experiment. S6 software can display 1D scatterplot plots, 2D plots, and 3D plots as shown in Fig. 8. To demonstrate the full capacities of the system, the study has performed the four-color fluorescence assay all one reaction, using a FAM-labelled probe targeting the M-BCR gene and a HEX-labelled probe targeting the M-BCR gene, a ROX-labeled probe targeting m-BCR and a Cy5-probe targeting the μ-BCR gene. As shown in Fig. 8A, all in a 1D plot, fluorescence values for a given detection channel, expressed as relative fluorescence units (RFU), are plotted on the vertical axis, whereas the values displayed on the horizontal axis represent the droplet index (Fig. 8A). In a 2D dot plot, fluorescence intensity values for one of the four detection channels are plotted on the vertical axis, while fluorescence intensity values for one of the other 2 detection channels are plotted on the horizontal axis (Fig. 8B). The use of these 2D scatterplots thus additionally allows the characterization of potential correlations in fluorescence for the different channels. A 3D dot plot (Fig. 8C) is the best way to capture all information at once for all four color fluorescent targets, however, it is not an easy format to manipulate for the human eye. In contrast, 1D and 2D dot plots can be more easily obtained, and can thus be used for further data analysis, such as threshold setting to discriminate negative from positive partitions.

Fig. 7 Visualization of droplets generated by C4 chips. Imaged post PCR using the FAM, HEX ROX, and CY5 four acquisition channels. The upper right corner magnifies the droplet morphology shown in the lower left corner presents a zoomed-in portion of the droplet crystal. Droplets in which no amplification has taken place remain dark, and droplets in which amplification is observed are lighter in color.

Fig. 8 S6 software can display 1D scatterplot plots, 2D plots, and 3D plots. (A) The one-dimensional plot generated by the digital PCR software, presenting four channels: FAM, HEX, ROX, and CY5. (B) The two dimensional plots for the FAM and HEX channels, as well as for the ROX and CY5 channels. (C) A captivating three dimensional plot is presented, combining data from the FAM, HEX, and CY5 channels, offering a more comprehensive and insightful representation of the experimental data.

Evaluate of quantitative accuracy

In the present study, ERM-AD623 was used in our quantitative study of BCR-ABL/ABL. The accuracy of the quantification was assessed by different concentrations of standards used and Bi was calculated for each target gene according to Bi = Mi − T. Bi calculated from the experimental results is in the range of −0.1 to 0.08, which is in the range of ±0.5. The associated experimental results are given in Table 3.

Table 3 The logarithms of the indicated values of the four ERM standards are 4.01, 3.01, 2.02, and 1.00, respectively. The Bi range of the BCR-ABL target is −0.1 to 0.05 and the ABL target is −0.07 to 0.08

Sample	BCR-ABL & ABL		BCR-ABL			ABL
	Labeled concentration copy per μL	T	Measured concentration copy per μL	Mi	Bi	Measured concentration copy per μL	Mi	Bi
ERM®-AD623c	10 300	4.01	8180	3.91	−0.10	8760	3.94	−0.07
ERM®-AD623d	1020	3.01	860	2.93	−0.07	870	2.94	−0.07
ERM®-AD623e	104	2.02	105.5	2.02	0.01	109.5	2.04	0.02
ERM®-AD623f	10	1.00	11.35	1.05	0.05	12.1	1.08	0.08

---

Evaluation of the limit of detection

In 15 CML-negative human blood samples and 15 different nucleic acid-free water samples, no positive copies of BCR-ABL were detected. The limit of blank (LOB) for the assay was set at 0. The evaluation of the limit of detection (LoD) was conducted using E14A2, E13A2, and ABL plasmids, and the LoD values were calculated using probit curves (refer to Fig. 9). The determined LoD for the assay is MR4.7 (0.002%).

Fig. 9 (A) LoD study of E14A2 and ABL plasmids. (B) LoD study of E13A2 and ABL plasmids. Digital PCR quantification of plasmid samples E14A2, E13A2, and ZL-ABL was performed in a low-concentration configuration with 32 replicates per concentration, and based on the results of the combined plasmid DNA study, the LoD of the assay can be determined as MR4.7 (0.002%).

Evaluation of the linearity

Linear studies were conducted on ERM standard products, in which the standard ERM®-AD623f was diluted twice to obtain the lowest concentration of the sample with a theoretical concentration of 5 copies per μL, a total of 5 concentrations. The results showed that BCR-ABL and the ABL target showed an excellent linear relationship (Fig. 10A and B). We further used plasmid samples ZL-E14A2 and ZL-E13A2 for linear studies, and the results also showed an excellent linear relationship (Fig. 10C and D). K562 cells and HL60 cells simulated negative samples, 50% IS samples, 10% IS samples, 1% IS samples, 0.1% IS samples, and 0.01% IS samples were used for linear analysis. Through the original droplet plots and one-dimensional droplet plots, the excellent linearity of our digital PCR detection system has been further verified (Fig. 11).

Fig. 10 Linear study results of the standard substance and plasmid samples E14A2 and E13A2. (A and B) A linear study was conducted on the standard, 5 concentrations were selected, and each concentration was measured twice. The results showed that the ERM standard of BCR-ABL and ABL indexes had an excellent linear relationship, and the calculated R² values were 0.9996 (A) and 0.9999 (B), respectively. (C and D) The linear study of sample E14A2 conducted and the linear study of sample E13A2 conducted also selected 5 concentrations and repeated measurements were made for each concentration three times, and the calculated R² values were 0.9906 (C) and 0.9913 (D), respectively, which also had an excellent linear relationship.

Fig. 11 The original droplet plot and one-dimensional plot exhibit the results of the linear evaluation through sample gradient dilution. The original droplet plot represents the signals of BCR-ABL1 and ABL targets on the fluorescence channels. The ABL gene serving as the internal reference maintains a stable copy number, while the BCR-ABL1 gene shows no amplification in the no-template control (NC). Within the IS range of 50% to 0.01%, the BCR-ABL1 gene demonstrates excellent linear variation.

Evaluation of precision

All three concentrations of the plasmid samples exhibited high precision in intra-day, inter-day, inter-instrument, and inter-lot assays (Table 4). Additionally, when the precision data were combined to calculate the total coefficient of variation, a high degree of precision was observed (Fig. 12). Our results indicate that the method used in this study has excellent precision for samples with different concentrations under different conditions (Table 2, Fig. 5).

Table 4 The results showed that all three samples had a CV of less than 4% for intra-day, inter-day, and inter-instrument variation. The CV of inter-batch variation was less than 7%. The total CV was less than 5%

Precision type	Sample MR value	Average MR values	SD	CV
Intra-day precision	MR2	1.959	0.054	2.74%
	MR3	3.008	0.06	2.01%
	MR4	3.942	0.101	2.55%
Inter-day precision	MR2	1.964	0.045	2.29%
	MR3	2.956	0.106	3.59%
	MR4	3.888	0.133	3.43%
Inter-instrument precision	MR2	1.976	0.06	3.02%
	MR3	3.011	0.056	1.88%
	MR4	3.954	0.13	3.29%
Inter-lot precision	MR2	1.954	0.056	2.87%
	MR3	2.894	0.196	6.78%
	MR4	3.769	0.213	5.66%
Total coefficient of variation	MR2	1.956	0.013	0.68%
	MR3	2.883	0.129	4.47%
	MR4	3.791	0.156	4.11%

---

Fig. 12 Samples MR2, MR3, and MR4 had 46, 46, and 47 replicates, respectively. The detected average MR values for these samples were 1.96, 2.88, and 3.79. The coefficients of variation for the three samples were 0.68%, 4.47%, and 4.11%, respectively.

Evaluation of the clinical sample

Bland–Altman plots were used to analyze the results of ddPCR and RT-qPCR assay methods (Fig. 6). The results showed that the mean bias between the two methods was −0.474, with a 95% confidence interval between −2.241 and 1.293 (Fig. 13). These findings suggest that there is excellent consistency between the two techniques for measuring the IS value in clinical samples. However, further studies are needed to determine which method is more accurate and reliable for this purpose.

Fig. 13 Measurement of the IS value in clinical samples was performed using qPCR and dPCR,^32,33 and the consistency of the two techniques was evaluated using Bland–Altman plots. The mean bias between the two methods was −0.474, with a standard deviation difference of 0.902 and a 95% confidence interval difference of −2.241 to 1.293.

Conclusion

In this paper, a new innovative and integrated digital drop PCR platform has been developed that harnesses cutting-edge microfluidic technology to integrate the dPCR workflow onto a single consumable chip.³⁴ This provides a fast and simple workflow with high multiplexing capability and powerful sensitivity, while also reducing the time to results. The entire experimental process takes less than 2.5 hours, making it easy to use. The whole process of droplet generation, PCR amplification, and droplet detection is completed on one chip, which meets the clinical requirement of “sample in, result out”. And it is equipped with four-color fluorescence to achieve multi-target simultaneous detection; the entire process is completely sealed through the use of an oil phase that is lighter than water. The whole process is completely closed, which can avoid cross-contamination between samples, and realize BCR-ABL fusion gene shear measurement and precise quantification. All our preliminary results throw light on the nature of the higher quantitative accuracy of dd-PCR in molecular detection of CML patients and the stability of PCR reaction inhibitors, as well as the development prospect of broad clinical application brought by its reproducibility.³⁵ These results show that the system has excellent practicability, and the data will be analyzed to draw further conclusions.

In this study, the new DropXpert S6 system was used by us for the first time. In terms of reliability and repeatability verification: 16 amplifications were performed on 4 chips on this platform, and the chips continuously produced about 25 000 analyzable droplets, showing a theoretical dynamic range of 5 orders of magnitude from 0.2 copies per μL to 20 000 copies per μL, which validates their exceptional reliability and repeatability; in terms of temperature uniformity and consistency verification: during the heating process, surfactants were used to measure the temperature at different points on the chip. At a given temperature of 95 °C and 60 °C, the average temperatures measured on the chip surface were 94.9 °C and 59.9 °C, respectively, indicating excellent consistency of temperature; in terms of accuracy evaluation: accuracy assessment by calculating the absolute bias of each target gene yielded a range from −0.1 to 0.08, all within ±0.5 logarithmic orders of magnitude; in terms of detection limit evaluation: 32 repetitions were performed at each concentration, the LOB for the assay was set at 0, and the LoD values were calculated using probit curves is MR4.7 (0.002%); in terms of linear evaluation: negative samples, 50% IS samples, 10% IS samples, 1% IS samples, 0.1% IS samples, and 0.01% samples were simulated for linear analysis using K562 cells and HL60 cells, the linearity evaluation showed that the R² value of the BCR-ABL was 0.9996, and the R² value of the ABL metrics calculated using the ERM standard was 0.9999, and the calculated R² values were 0.9906 and 0.9913 for plasmid samples ZL-E14A2 and ZL-E13A2, respectively. In terms of precision evaluation: two experiments were performed every day, one in the morning and one in the afternoon, two instruments were used for each experiment, and each concentration sample was repeated 8 times, batch 1 was performed for 2 days, and batch 2 was performed for 1 day, a total of 12 chips and 48 repetitions, and the precision evaluation showed that all samples had a CV of less than 4% for intra-day, inter-day, and inter-instrument variation. The CV of inter-batch variation was less than 7%. The total CV was less than 5%; in terms of clinical sample consistency: all measurements were within the 95% confidence interval. The experimental results show that the new DropXpert S6 system has sufficient scientific basis and stability for the status certification of CML patients, which is conducive to the detection and treatment of CML patients in the future.

Author contributions

Wenjia Wei: data curation, writing – original draft. Shujun Li and Qun He: investigation. Linfang Yu: data curation, methodology, and supervision. Junwei Ye: formal analysis, validation, visualization, and writing – original draft. Ying Zhang, Simin Deng, Weiwei Zhao, Xielan Zhao, and Yajing Xu: data curation and validation. Zhiping Jiang: conceptualization, funding acquisition, project administration, resources, review & editing and supervision.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This study was supported by the Key Research and Development Plan Projects in Hunan Province of China (Grant No. 2022SK2005) and the Natural Science Foundation of Hunan Province of China (Grant No. 2023JJ30895).

References

1. K. Shinjo, K. Hara, G. Nagae, T. Umeda, K. Katsushima and M. Suzuki, et al., A novel sensitive detection method for DNA methylation in circulating free DNA of pancreatic cancer, PLoS One, 2020, 15(6), e0233782.
2. E. A. Ottesen, J. W. Hong, S. R. Quake and J. R. Leadbetter, Microfluidic digital PCR enables multigene analysis of individual environmental bacteria, Science, 2006, 314(5804), 1464–1467.
3. B. J. Hindson, K. D. Ness and D. A. Masquelier, et al., High-throughput droplet digital PCR system for absolute quantitation of DNA copy number[J], Anal. Chem., 2011, 83(22), 8604–8610.
4. M. Baker, Digital PCR hits its stride[J], Nat. Methods, 2012, 9(6), 541–544.
5. S. Soverini, S. De Santis, M. Martelli, C. Monaldi, F. Castagnetti, G. Gugliotta, C. Papayannidis, M. Mancini, S. Bruno, C. Venturi, K. Machova Polakova, T. Ernst, D. Maar, A. Corner and M. Cavo, Droplet digital PCR for the detection of second-generation tyrosine kinase inhibitor-resistant BCR::ABL1 kinase domain mutations in chronic myeloid leukemia, Leukemia, 2022, 36(9), 2250–2260,  DOI:10.1038/s41375-022-01660-8 , Epub 2022 Jul 30. PMID: 35908105.
6. G. Colafigli, E. Scalzulli, A. Di Prima, S. Pepe, M. G. Loglisci, D. Diverio, M. Martelli, R. Foà and M. Breccia, Digital droplet PCR as a predictive tool for successful discontinuation outcome in chronic myeloid leukemia: Is it time to introduce it in the clinical practice?, Crit. Rev. Oncol. Hematol., 2021, 157, 103163,  DOI:10.1016/j.critrevonc.2020.103163 , Epub 2020 Nov 12. PMID: 33246263.
7. P. Liao and Y. Huang, Digital PCR: Endless Frontier of 'Divide and Conquer', Micromachines, 2017, 8(8), 231.
8. J. W. Vardiman, Chronic Myelogenous Leukemia,BCR-ABL1+, Am. J. Clin. Pathol., 2009, 132, 250–260.
9. J. D. Rowley, Letter: A new consistent chromosomal abnormality in chronic myelogenous leukaemia identified by quinacrine fluorescence and Giemsa staining, Nature, 1973, 243(5405), 290–293,  DOI:10.1038/243290a0 , PMID: 4126434.
10. G. Kaur, V. L. Narayanan, P. A. Risbood, M. G. Hollingshead, S. F. Stinson, R. K. Varma and E. A. Sausville, Synthesis, structure-activity relationship, and p210(bcr-abl) protein tyrosine kinase activity of novel AG 957 analogs, Bioorg. Med. Chem., 2005, 13(5), 1749–1761,  DOI:10.1016/j.bmc.2004.12.003 , PMID: 15698792.
11. O. Kalinina, I. Lebedeva, J. Brown and J. Silver, Nanoliter scale PCR with TaqMan detection, Nucleic Acids Res., 1997, 25(10), 1999–2004,  DOI:10.1093/nar/25.10.1999 , PMID: 9115368; PMCID: PMC146692.
12. P. J. Sykes, S. H. Neoh, M. J. Brisco, E. Hughes, J. Condon and A. A. Morley, Quantitation of targets for PCR by use of limiting dilution, BioTechniques, 1992, 13(3), 444–449 , PMID: 1389177.
13. B. Vogelstein and K. W. Kinzler, Digital PCR, Proc. Natl. Acad. Sci. U. S. A., 1999, 96(16), 9236–9241,  DOI:10.1073/pnas.96.16.9236 , PMID: 10430926; PMCID: PMC17763.
14. D. Svec, A. Tichopad, V. Novosadova, M. W. Pfaffl and M. Kubista, How good is a PCR efficiency estimate: Recommendations for precise and robust qPCR efficiency assessments, Biomol. Detect. Quantif., 2015, 3, 9–16,  DOI:10.1016/j.bdq.2015.01.005 , PMID: 27077029; PMCID: PMC4822216.
15. P. A. Bartley, D. M. Ross, S. Latham, M. H. Martin-Harris, B. Budgen, V. Wilczek, S. Branford, T. P. Hughes and A. A. Morley, Sensitive detection and quantification of minimal residual disease in chronic myeloid leukaemia using nested quantitative PCR for BCR-ABL DNA, Int. J. Lab. Hematol., 2010, 32(6 Pt 1), e222–e228,  DOI:10.1111/j.1751-553X.2010.01236.x , PMID: 20491997.
16. W. J. Wang, C. F. Zheng, Z. Liu, Y. H. Tan, X. H. Chen, B. L. Zhao, G. X. Li, Z. F. Xu, F. G. Ren, Y. F. Zhang, J. M. Chang and H. W. Wang, Droplet digital PCR forCML (P210) detection of chronic myeloid leukemia: A high sensitive method of the minimal residual disease and disease progression, Eur. J. Haematol., 2018, 101(3), 291–296,  DOI:10.1111/ejh.13084 , Epub 2018 Jul 12. PMID: 29691899.
17. G. Cathomas, C. McGandy, I. Diamantis, A. Tichelli and F. Gudat, Nachweis von bcr-abl-Transkripten in “minimal residual disease” bei chronischer myeloischer Leukämie mittels der nested polymerase chain reaction (PCR) [Detection of bcr-abl transcripts in “minimal residual disease” in chronic myelogenous leukemia by nested polymerase chain reaction (PCR)], Verh. Dtsch. Ges. Pathol., 1994, 78, 314–317 , German. PMID: 7534004.
18. J. E. Cortes, D. Jones, S. O'Brien, E. Jabbour, F. Ravandi, C. Koller, G. Borthakur, B. Walker, W. Zhao, J. Shan and H. Kantarjian, Results of dasatinib therapy in patients with early chronic-phase chronic myeloid leukemia, J. Clin. Oncol., 2010, 28(3), 398–404,  DOI:10.1200/JCO.2009.25.4920 , Epub 2009 Dec 14. PMID: 20008620; PMCID: PMC2815702.
19. K. Sasaki, S. S. Strom, S. O'Brien, E. Jabbour, F. Ravandi, M. Konopleva, G. Borthakur, N. Pemmaraju, N. Daver, P. Jain, S. Pierce, H. Kantarjian and J. E. Cortes, Relative survival in patients with chronic-phase chronic myeloid leukaemia in the tyrosine-kinase inhibitor era: analysis of patient data from six prospective clinical trials, Lancet Haematol., 2015, 2(5), e186–e193,  DOI:10.1016/S2352-3026(15)00048-4 , Epub 2015 Apr 20. PMID: 26688093; PMCID: PMC4884053.
20. M. Baccarani, M. W. Deininger, G. Rosti, A. Hochhaus, S. Soverini, J. F. Apperley, F. Cervantes, R. E. Clark, J. E. Cortes, F. Guilhot, H. Hjorth-Hansen, T. P. Hughes, H. M. Kantarjian, D. W. Kim, R. A. Larson, J. H. Lipton, F. X. Mahon, G. Martinelli, J. Mayer, M. C. Müller, D. Niederwieser, F. Pane, J. P. Radich, P. Rousselot, G. Saglio, S. Saußele, C. Schiffer, R. Silver, B. Simonsson, J. L. Steegmann, J. M. Goldman and R. Hehlmann, European LeukemiaNet recommendations for the management of chronic myeloid leukemia: 2013, Blood, 2013, 122(6), 872–884,  DOI:10.1182/blood-2013-05-501569 , Epub 2013 Jun 26. PMID: 23803709; PMCID: PMC4915804.
21. A. Hochhaus, M. Baccarani, R. T. Silver, C. Schiffer, J. F. Apperley, F. Cervantes, R. E. Clark, J. E. Cortes, M. W. Deininger, F. Guilhot, H. Hjorth-Hansen, T. P. Hughes, J. J. W. M. Janssen, H. M. Kantarjian, D. W. Kim, R. A. Larson, J. H. Lipton, F. X. Mahon, J. Mayer, F. Nicolini, D. Niederwieser, F. Pane, J. P. Radich, D. Rea, J. Richter, G. Rosti, P. Rousselot, G. Saglio, S. Saußele, S. Soverini, J. L. Steegmann, A. Turkina, A. Zaritskey and R. Hehlmann, European LeukemiaNet 2020 recommendations for treating chronic myeloid leukemia, Leukemia, 2020, 34(4), 966–984,  DOI:10.1038/s41375-020-0776-2 , Epub 2020 Mar 3. PMID: 32127639; PMCID: PMC7214240.
22. R. Hehlmann, Chronic Myeloid Leukemia in 2020, HemaSphere, 2020, 4(5), e468,  DOI:10.1097/HS9.0000000000000468 , PMID: 33134861; PMCID: PMC7544304.
23. F. X. Mahon and G. Etienne, Clin. Cancer Res., 2014, 20(2), 310–322,  DOI:10.1158/1078-0432.CCR-13-1988 , Epub 2013 Oct 28. PMID: 24166905.
24. T. P. Hughes and D. M. Ross, Moving treatment-free remission into mainstream clinical practice in CML, Blood, 2016, 128(1), 17–23,  DOI:10.1182/blood-2016-01-694265 , Epub 2016 Mar 24. PMID: 27013442.
25. F. X. Mahon, D. Réa, J. Guilhot, F. Guilhot, F. Huguet, F. Nicolini, L. Legros, A. Charbonnier, A. Guerci, B. Varet, G. Etienne, J. Reiffers, P. Rousselot and Intergroupe Français des Leucémies Myéloïdes Chroniques, Discontinuation of imatinib in patients with chronic myeloid leukaemia who have maintained complete molecular remission for at least 2 years: the prospective, multicentre Stop Imatinib (STIM) trial, Lancet Oncol., 2010, 11(11), 1029–1035,  DOI:10.1016/S1470-2045(10)70233-3 , Epub 2010 Oct 19. PMID: 20965785.
26. P. L. Quan, M. Sauzade and E. Brouzes, dPCR: A Technology Review, Sensors, 2018, 18(4), 1271,  DOI:10.3390/s18041271 , PMID: 29677144; PMCID: PMC5948698.
27. J. Cortes, C. Pavlovsky and S. Saußele, Chronic myeloid leukaemia, Lancet, 2021, 398(10314), 1914–1926,  DOI:10.1016/S0140-6736(21)01204-6 , Epub 2021 Aug 20. PMID: 34425075.
28. C. L. Sawyers, J. McLaughlin, A. Goga, M. Havlik and O. Witte, The nuclear tyrosine kinase c-Abl negatively regulates cell growth, Cell, 1994, 77(1), 121–131,  DOI:10.1016/0092-8674(94)90240-2 , PMID: 7512450.
29. E. Jabbour and H. Kantarjian, Chronic myeloid leukemia: 2022 update on diagnosis, therapy, and monitoring, Am. J. Hematol., 2022, 97(9), 1236–1256,  DOI:10.1002/ajh.26642 , Epub 2022 Jul 6. PMID: 35751859.
30. W. Wang, J. E. Cortes, G. Tang, J. D. Khoury, S. Wang, C. E. Bueso-Ramos, J. A. DiGiuseppe, Z. Chen, H. M. Kantarjian, L. J. Medeiros and S. Hu, Risk stratification of chromosomal abnormalities in chronic myelogenous leukemia in the era of tyrosine kinase inhibitor therapy, Blood, 2016, 127(22), 2742–2750,  DOI:10.1182/blood-2016-01-690230 , Epub 2016 Mar 22. PMID: 27006386; PMCID: PMC4915795.
31. D. J. Barnes, D. Palaiologou, E. Panousopoulou, B. Schultheis, A. S. Yong, A. Wong, L. Pattacini, J. M. Goldman and J. V. Melo, Bcr-Abl expression levels determine the rate of development of resistance to imatinib mesylate in chronic myeloid leukemia, Cancer Res., 2005, 65(19), 8912–8919,  DOI:10.1158/0008-5472.CAN-05-0076 , PMID: 16204063.
32. M. Baccarani, J. Cortes, F. Pane, D. Niederwieser, G. Saglio, J. Apperley, F. Cervantes, M. Deininger, A. Gratwohl, F. Guilhot, A. Hochhaus, M. Horowitz, T. Hughes, H. Kantarjian, R. Larson, J. Radich, B. Simonsson, R. T. Silver, J. Goldman, R. Hehlmann and European LeukemiaNet, Chronic myeloid leukemia: an update of concepts and management recommendations of European LeukemiaNet, J. Clin. Oncol., 2009, 27(35), 6041–6051,  DOI:10.1200/JCO.2009.25.0779 , Epub 2009 Nov 2. PMID: 19884523; PMCID: PMC4979100.
33. N. Testoni, G. Marzocchi, S. Luatti, M. Amabile, C. Baldazzi, M. Stacchini, M. Nanni, G. Rege-Cambrin, E. Giugliano, U. Giussani, E. Abruzzese, S. Kerim, M. G. Grimoldi, A. Gozzetti, B. Crescenzi, C. Carcassi, P. Bernasconi, A. Cuneo, F. Albano, G. Fugazza, A. Zaccaria, G. Martinelli, F. Pane, G. Rosti and M. Baccarani, Chronic myeloid leukemia: a prospective comparison of interphase fluorescence in situ hybridization and chromosome banding analysis for the definition of complete cytogenetic response: a study of the GIMEMA CML WP, Blood, 2009, 114(24), 4939–4943,  DOI:10.1182/blood-2009-07-229864 , Epub 2009 Oct 1. PMID: 19797518.
34. J. W. Hong and S. R. Quake, Integrated nanoliter systems, Nat. Biotechnol., 2003, 21(10), 1179–1183,  DOI:10.1038/nbt871 , PMID: 14520403.
35. R. J. Martinez, Q. Kang, D. Nennig, N. G. Bailey, N. A. Brown, B. L. Betz, M. Tewari, B. Thyagarajan, V. Bachanova and P. Mroz, One-Step Multiplexed Droplet Digital Polymerase Chain Reaction for Quantification of p190 BCR-ABL1 Fusion Transcript in B-Lymphoblastic Leukemia, Arch. Pathol. Lab. Med., 2022, 146(1), 92–100,  DOI:10.5858/arpa.2020-0454-OA , PMID: 33769465.

---