Advertisement

Establishment of low-cost laboratory automation processes using AutoIt and 4-axis robots

Open AccessPublished:July 10, 2022DOI:https://doi.org/10.1016/j.slast.2022.07.001

      Abstract

      In most small laboratories, many processes are not yet automated because existing laboratory automation solutions are usually expensive and inflexible to use. Examples of this are autosamplers that are only compatible with one specific laboratory instrument or larger liquid handling stations that are expensive and usually self-contained. A flexible and inexpensive way to automate laboratory processes would be to automate existing laboratory equipment with the help of suitable robotic arms. In this study, we investigate the feasibility of such a strategy based on a low-cost 4-axis robot and freely available software. We used the scripting language AutoIt that automates any Windows-based instrument control software. Using these tools, we automated three fundamentally different laboratory processes: a pipetting process, a use as an autosampler for an atomic absorption spectroscopy instrument, and a more complex process involving the inoculation of bacterial cultures. We also integrated a conventional webcam for 2D barcode recognition. Compared to a trained professional who performed all experiments manually, all setups showed no significant differences in accuracy and precision. In summary, the tested system consisting of a 4-axis robot and freely available software is suitable for flexible automation and has potential for even more complex laboratory processes. Limitations such as a lack of collaboration and speed will be addressed in follow-up studies. The system thus represents a well-suited flexible laboratory automation system for both research and teaching purposes.

      Keywords

      Abbreviations:

      AAS (Atomic absorption spectroscopy), DIY (Do-it-yourself), ELISA (Enzyme-linked immunosorbent assay), GUI (Graphical user interface), OPC LADS (Open Platform Communications Laboratory Agnostic Device Standard), PLA (Polylactid acid), SiLA (Standardization in Lab Automation), SME (Small-medium-enterprise)

      Introduction

      Highly professional solutions for the automation of laboratory processes have existed for decades. In certain areas, such as clinical analysis, this has not only led to an enormous increase in work productivity and throughput, but has also made a decisive contribution to the high quality of the bioanalytical results [
      • Thomson Jr., R.B.
      • McElvania E.
      Total laboratory automation: what is gained, what is lost, and who can afford it?.
      ,
      • Lippi G.
      • Da Rin G.
      Advantages and limitations of total laboratory automation: a personal overview.
      ]. The investment costs for such laboratory lines can be in excess of 1 million € [

      Personal communication, Siemens Healthcare.

      ].
      Thus, especially in the small-medium-enterprise (SME) sector and in small research laboratories, often only very few laboratory processes are automated. This means that many of the advantages of laboratory automation (quality improvements, productivity increases, etc.) are lost to this sector.
      For SMEs, the laboratory automation market often offers isolated solutions such as liquid-handling platforms that cover individual tasks—such as performing immunoassays—very well [
      • Neubauer J.C.
      • Sébastien I.
      • Germann A.
      • et al.
      Towards a full automation of the elispot assay for safe and parallelized immunomonitoring.
      ]. However, investment costs of up to approximately 1 million € (and follow-up costs such as for maintenance) deter biotech-SMEs from such laboratory automation projects. Furthermore, the introduction of such laboratory automation solutions often means that existing, expensive, manually operated instruments are no longer in use. Commercially offered laboratory automation solutions are usually highly specialized devices that can only be used for a few analytical techniques at a time. However, flexibility is required, especially in the SME sector, as they mostly perform project-related, temporary activities. These automation solutions should ideally be adaptable to new project requirements in a flexible way, or the high investment costs will not pay off. For the sandwich enzyme-linked immunosorbent assay (ELISA), for example, a suitable liquid-handling system would have to be purchased. Apart from the high investment costs, this system would be unsuitable elsewhere in the laboratory, such as for automated sample feeding for a chromatography system. Of course, low-cost solutions for automation of single laboratory tasks do exist, as is the case for ELISA applications [
      • Mohammed M.I.
      • Desmulliez M.P.Y
      Lab-on-a-chip based immunosensor principles and technologies for the detection of cardiac biomarkers: a review.
      ,
      • Wang G.
      • Das C.
      • Ledden B.
      • et al.
      Development of fully automated low-cost immunoassay system for research applications.
      ]. However, a flexible and inexpensive laboratory robot system that can be used for a wide variety of laboratory processes and without significant effort does not yet exist but would be desirable. One obstacle to resolve this challenge is the fact that interoperability between laboratory devices is hardly available or at most within a product range of a specific manufacturer. This applies to control as well as to data exchange between components.
      The Standardization in Lab Automation (SiLA) Consortium addresses this problem by introducing free and open system communication and data standards across different laboratory automation platforms [,
      • Bär H.
      • Hochstrasser R.
      • Papenfuß B.
      SiLA: basic standards for rapid integration in laboratory automation.
      ]. Although a number of manufacturers have now implemented the SiLA standards, there are still many devices that are not SiLA compliant [
      • Bär H.
      • Hochstrasser R.
      • Papenfuß B.
      SiLA: basic standards for rapid integration in laboratory automation.
      ,
      • Carvalho M.C.
      Practical Laboratory Automation: Made easy with AutoIt.
      ]. Older, existing laboratory equipment is especially difficult to integrate in automated processes. Another initiative to standardize communication between lab devices is Open Platform Communications Laboratory Agnostic Device Standard (OPC LADS), which is still in an early stage [].
      The market for laboratory automation solutions currently offers highly professional, but also very expensive and inflexible systems, which are either suitable only for the automation of single instruments (e.g., so-called autosamplers {approximately 10K-25K €}), large-scale solutions such as entire laboratory lines {>1 million €} or self-contained liquid handling systems {approximately 10K - 300K €} [
      • Thomson Jr., R.B.
      • McElvania E.
      Total laboratory automation: what is gained, what is lost, and who can afford it?.
      ,
      • Lippi G.
      • Da Rin G.
      Advantages and limitations of total laboratory automation: a personal overview.
      ,
      • Carvalho M.C.
      • Eyre D.E.
      A low cost, easy to build, portable, and universal autosampler for liquids.
      ].
      In addition, there are a few low-cost laboratory automation solutions on the market (such as Andrew+ or Opentrons), but these do not solve the problem of existing laboratory devices such as thermocyclers, incubators or shakers that cannot be integrated into the automation system [
      • Reed C.E.
      • Fournier J.
      • Vamvoukas N.
      • et al.
      Automated preparation of ms-sensitive fluorescently labeled N-glycans with a commercial pipetting robot.
      ,
      • Zucchelli P.
      • Horak G.
      • Skinner N.
      ,
      • Councill E.E.A.W.
      • Axtell N.B.
      • Truong T.
      • et al.
      Adapting a low-cost and open-source commercial pipetting robot for nanoliter liquid handling.
      ,
      • Storch M.
      • Haines M.C.
      • Baldwin G.S.
      DNA-BOT: a low-cost, automated DNA assembly platform for synthetic biology.
      ,
      • Eggert S.
      • Mieszczanek P.
      • Meinert C.
      • Hutmacher D.W.
      ]. Although these systems—especially Opentrons compared to the previous version—have gained in user-friendliness, the systems require the purchase of manufacturer specific laboratory devices, even though such devices from other vendors already exist in the laboratory.
      Also, the idea of low-cost do-it-yourself (DIY) laboratory automation is, of course, not new. Examples include the already mentioned immunoassay applications and a number of other applications like toxicological bioassays, nucleic acid isolation and amplification or microfluidic protein arrays [
      • Steffens S.
      • Nüßer L.
      • Seiler T.B.
      • et al.
      A versatile and low-cost open source pipetting robot for automation of toxicological and ecotoxicological bioassays.
      ,
      • Chan K.
      • Koen M.
      • Hardick J.
      • et al.
      Low-cost 3D printers enable high-quality and automated sample preparation and molecular detection.
      ,
      • Dixit C.K.
      • Kadimisetty K.
      • Otieno B.A.
      • et al.
      Electrochemistry-based approaches to low cost, high sensitivity, automated, multiplexed protein immunoassays for cancer diagnostics.
      ].
      However, this requires that employees have specific qualifications in for example programming and a certain kind of engineering and thereby have an affinity for laboratory automation. If an automation system could handle multiple tasks in the lab, employees would at least only have to deal with one system and have greater benefits at the same time.
      In this study, we tried to establish a flexible, yet inexpensive laboratory automation system that has the potential to work with a wide range of laboratory instruments from different manufacturers and fulfill various laboratory tasks. As a proof-of-principle, three laboratory processes were automated: a pipetting process, an autosampler for an atomic absorption spectroscopy (AAS) device and the inoculation of bacterial cultures.
      We used an inexpensive 4-axis robot (Dobot Magician) for the physical tasks of the automated processes. The robotic arm provides a simple control software by the manufacturer and was already used in a medical laboratory in the field of sample preparation [,
      • Ciulu M.D.
      • Stoicu-Tivadar L.
      • Benis A.
      Automating processes in laboratories with the support of dobot magician.
      ].
      For the different laboratory tasks like gripping, pipetting and transfer of objects, the robot arm had to be equipped with different end effectors. 3D printing has proven to be an often-used technology to develop necessary parts for do-it-yourself lab automation systems [
      • Zhang C.
      • Wijen B.
      • Pearce J.M.
      Open-source 3-D platform for low-cost scientific instrument ecosystem.
      ,
      • Barthels F.
      • Berthels U.
      • Schwickert M.
      • Schirmeister T.
      FINDUS: an open-source 3D printable liquid-handling workstation for laboratory automation in life sciences.
      ]. In this study we either applied end effectors provided by the manufacturer, or developed others like a pipet holder with 3D printing. Some robotic arms are equipped with different sensors to automatically interact with their environment. Although these kinds of sensors extend the range of functions, they also significantly increase the cost of such systems. We decided to include a simple and low-cost, yet effective, barcode recognition function into our system, based on a commercial webcam and a suitable recognition software.
      For the simultaneous control of the different laboratory instruments and the robot arm, a user interface needed to be established. This user interface should be able to not only control laboratory instruments and the robot arm, but also all relevant software packages used in a laboratory process. Ideally, this user interface can operate the user windows on a Windows computer in an automated fashion in the same way as a human user. We used AutoIt, a scripting tool that is freely available, and that can control the graphical user interface (GUI) of any Windows operational system from Windows XP to Windows 10 []. This means that AutoIt can control any Windows software via its normal GUI—much like a user would do—using for example mouse clicks, shortcuts and more complex commands. AutoIt as a tool has been successfully used several times before in the field of self-made laboratory automation as described in numerous publications [
      • Carvalho M.C.
      • Murray R.H.
      Osmar, the open-source microsyringe autosampler.
      ,
      • Fleischer H.
      • Quang Do V.
      • Thurow K.
      Online measurement system in reaction monitoring for determination of structural and elemental composition using mass spectrometry.
      ,
      • Carvalho M.C.
      ,
      • Carvalho M.
      Integration of analytical instruments with computer scripting.
      ].
      A precise and extensive description of this scripting language and its possibilities in the laboratory are described by Carvalho, M. C. (2016) in ”Practical Laboratory Automation Made it easy with AutoIt” [].

      Materials and methods

      For all automation projects, we utilized the 4-axis robot system Dobot Magician (DOBOT, Shenzhen, China) as main hardware component, which is sold as a training robotic arm [
      • Dixit C.K.
      • Kadimisetty K.
      • Otieno B.A.
      • et al.
      Electrochemistry-based approaches to low cost, high sensitivity, automated, multiplexed protein immunoassays for cancer diagnostics.
      ]. We also used the corresponding DobotStudio V1.9.2 software (Variobotic GmbH, Neu-Ulm, Germany) to program its motion sequences. The system is sold for approximately 1.5K €, has a maximum payload of 0.5 kg, a maximum reach of 320 mm and is advertised with a position repeatability of 0.2 mm. Other specifications can be viewed at https://www.dobot.cc/dobot-magician/specification.html. Several end effectors like a gripper and a suction cup are provided in the system. A linear rail system can be additionally acquired to extend the maximum range by 1000 mm (linear extension).
      We applied the scripting tool AutoIt (version 3.3.14.5, AutoIt Consulting Ltd., Worcestershire, UK) to manage all software-based automation tasks. The software is freely available. AutoIt can control the GUI of any Windows operational system from Windows XP to Windows 10 []. The AutoIt script contains particular variables, conditional loops and nested if-functions. In addition, we created GUIs as an interface between the application and the user. These GUIs were designed with the help of the KODA Form Designer, which is included in the AutoIt installation package. The installation package and more information about the scripting language AutoIt can be found at https://www.autoitscript.com/site/.

      Pipetting process

      We mounted the Dobot Magician on its linear axis to increase the radius of action for the automated pipetting process and used a self-designed pipet holder to gain the pipet function. This pipet holder was 3D printed from PLA (polylactid acid, racemic-atactic) and can hold any single channel electronic micropipet from the manufacturer BRAND (BRAND GmbH & Co. KG, Wertheim, Germany) in the range from 0.1 µL to 5000 µL. It operates through a linear solenoid actuator to press the main pipetting button of the chosen pipet for aspirating and dispensing. Both, the pipet movement and the execution of liquid pipetting were controlled by DobotStudio running on Windows 10.
      We demonstrated the accuracy of automated pipetting performing two different experimental setups. In these experiments, we utilized the 200 µL Transferpette (BRAND) with related pipet tips (200 µL, BRAND) to transfer different concentrations of lissamine green (576.62 g/mol, CARL ROTH GmbH & Co. KG, Karlsruhe, Germany) diluted in deionized water (15 MΩ).
      We tested the multi pipetting function in the first experimental setup, in which we prepared the wells of 96-well target plate (CARL ROTH GmbH & Co. KG, Karlsruhe, Germany) with 200 µL deionized water (15 MΩ). While the automated process, the pipet aspired 160 µL of diluent dispensed it in 20 µL steps into the target plate. Each dilution (blank, 0.048 mg/mL and 0.06 mg/mL lissamine green dilution) were pipetted in two columns [n = 16]. Before liquid change, the pipet tip was rinsed in a vessel of deionized water automatically.
      In the second experiment, we tested the function of single pipetting. The system pipetted 100 µL each of the blank solution, 0.003 mg/µL, 0.006 mg/mL, 0.012 mg/mL and 0.048 mg/mL of the lissamine green dilution into the 96-well plate sequentially and in columns [n=12], and rinsed the pipet tip automatically with deionized water after each column. Trained laboratory staff performed both experiments (multi and single pipetting setup) with pipet tip change after each dilution change. Absorbance of all 96-well target plates were measured at 635 nm by SpectraMax iD5 spectrometer (Molecular Device LLC, San Jose, California). We evaluated statistical differences with the Mann-Whitney U tests and set the significance threshold at p=0.05.

      Autosampler

      The AAS (AA240FS, Varian, today 240FS AA, Agilent, Santa Clara, CA) instrument has a sample capillary which is connected to the instrument via a plastic tube. The sample capillary must be immersed in a sample liquid so that the sample is automatically aspirated into the instrument for measurement with the aid of a pump on the AAS instrument. For the Autosampler we used the multi-tool adapter (Variobotic GmbH, Neu-Ulm, Germany), which has two vertical borings for various applications. One of those borings functioned as holder for the sampling capillary. We mounted the Dobot Magician—that holds the sampling capillary vertical—on a working plate together with racks for wash and sample containers. The function of AutoIt was to coordinate the workflow between the robot and the AAS measurements (controlled by the AAS software SpectrAA Version 5.1 Pro, running on a Windows XP) and to ensure communication with the user. We organized our validation experiment in four rounds (each manually and automatically) and performed it according to DIN EN 1134 (method L31.00-10)[]. In each round we measured Ca2+ and Mg2+ concentration in four vessels with ultrapure water (18 MΩ, deionized water processed with Milli-Q Gradient A10 with Quantum EX and Q-Gard 2, Merck Millipore, Burlington, Massachusetts) [n = 16], two vessels each with 2.5 mg/L standard [n=8] and 7.5 mg/L standard [n=8], and three vessels with sample of cherry juice [n=12]. Between each vessel, the sampling capillary was cleaned with a tissue in the manual process and by a series of automated washing steps (rinsed with deionized water, 15 MΩ) in the automated process. For statistical analysis, we performed hypothesis test (Wilcoxon Test, threshold 5%) for each element and sample.

      Bacterial inoculation

      For bacterial inoculation, the robot should fulfill three functions: first, removing and placing the plate lids; second, identifying the plates by barcodes; and third, inoculating the bacteria itself. We chose the suction cup provided by the robot manufacturer as end effector, to transfer the plate lids by negative pressure. On the suction cup, we fixed standard webcam LifeCam HD-3000 (Microsoft, Redmond, Washington) with its mount for realizing the second function. We enabled the robot to perform the actual bacterial inoculation by attaching an inoculation loop to an existing thread of the end effector and bending the inoculation loop downward by 90° at the front. The robot arm with its end effectors were mounted on a self-designed stainless steel plate (alloy 1.4301/X5CrNi18-10) and positioned within a laminar flow bench. The stainless steel plate also provided placeholders for three plates (one source plate and a maximum of two target plates) and their lids. The carrier plate can be disinfected with standard laboratory disinfectants. A gas burner was placed next to the carrier plate. We wrote an AutoIt script for creating a user interface and for the automatic coordination of Dobot Studio with the barcode recognition software bcWebCam (QS QualitySoft, Version 2.4.0.21, running on Windows 10). We compared inoculation success regarding cross contamination and inoculation quantity by inoculating for 12 well plates (Sarstedt AG & Co. KG, Nümbrecht, Germany) both manually and automatically. To prepare the bacterial suspension for the source plate, culture medium (casein soya bean digest broth, sifin diagnostics, GmbH, Berlin, Germany) was inoculated with a single colony of Escherichia coli (strain K12 RR28, DSM 4415) and incubated overnight in a centrifuge tube on a shaker (37°C, 180 rpm, type 3031, GFL mbH, Burgwedel, Germany). After a centrifugation (Eppendorf SE, Hamburg, Germany) step (4°C, 6.000 rpm, 4629 g, 5 min), we resuspended the bacterial pellet with 7 mL culture medium and 7 mL of 80% glycerol solution (Th. Geyer GmbH & Co. KG, Renningen, Germany) to get a 1:1 ratio. As a negative control, another test tube with a 1:1 mixture of culture medium and 80% glycerol solution was prepared.
      We alternately pipetted sample and negative control into the wells to check the sterilization process and sterile mode of operation after the experiment. Here we filled the wells of a 12-well plate (source plate) with 2 mL of either bacterial suspension or the negative control. For preparing the target plates, we pipetted 1 mL culture medium into each well of the target plates. In the automated process, the inoculation loop was sterilized for 4 s by a gas burner after each inoculation step, while we used 1 µL one-way inoculation loops (Th. Geyer GmbH & Co. KG) in the manual process. First, we did the experiment manually, then automated to see if the automation system causes cross-contaminations in source or target plates. Bacterial growth was assessed via a Tecan Sunrise spectrometer (Tecan Group, Männedorf, Switzerland) (cumulative absorbance at 600 nm, n=3) after incubation (24 h, 37°C). We compared the resulting data statistically by non-parametric Levene test (significance threshold p=0.05).

      Results and discussion

      As one of the first step towards a flexible and inexpensive laboratory automation system, we had to establish a strategy that would allow us to potentially connect a large variety of different, existing laboratory instruments with each other. The aim was not only to connect instruments from various manufacturers with each other, but also to include instruments in the automation process, which do not even have an interface for PC-connection (e.g., RS232 or USB-Port) like a stand-alone shaker, for example. To automate different laboratory processes with the 4-axis robot Dobot Magician, we utilized, on the one hand, the provided tools from the robot manufacturer, on the other hand, we created a 3D printed pipet holder.
      As a next step, a suitable AutoIt-based control software needed to be established to address two aims: first, a suitable user interface needed to be created. Second, a set of control routines needed to be prepared within this software, flexible enough to allow a variety of different automation processes in the laboratory.
      We included some of these routines in a universal control script that was subsequently used for each of the three demonstrated automation projects below. This universal control script included subroutines like GUIs for the startup process, safety dialogues, calibration procedures, general features like error responses or functions like the automated monitoring of a desired pixel area on the screen. The universal control script was supplemented with an additional script designed for each of the three automation projects below.
      We used this strategy in all three automation processes: AutoIt as a user interface and control software to both operate the robot control software “DobotStudio” and additional third-party software packages as required in the process (Fig. 1).
      Fig. 1
      Fig. 1System Environment of three automated laboratory processes. The software AutoIt was used as a superordinate system that provides the user interface and controls and coordinates the third-party software: DobotStudio, SpectrAA and bcWebCam. DobotStudio controls the desktop robotic arm for a) automated pipetting, b) for the autosampler of the atomic absorption spectroscopy instrument (AAS, b), and c) for automated bacteria inoculation. The SpectrAA software controls the AAS and records the data. The third-party software bcWebCam reads barcodes on the labware scanned by the webcam, installed on the robotic arm.
      The initial work to set up the control software included significant scripting. Fig. S1 shows a short example of such a script. The example also shows some corresponding dialogue boxes as GUIs with which the user interacts in the lab. As mentioned above, several generalized functions were included in the universal control script. This could be used in each of the three automation projects and required no further adaptations. However, additional scripts were required to adapt AutoIt to each automation project.
      While AutoIt is easy to learn, there is no doubt that not every researcher in the laboratory is familiar with this kind of scripting []. Nevertheless, in the courses at our university, we regularly observe that even untrained students in their second year are capable of using AutoIt for smaller automation projects after only 10 hours of training.
      After the general setup was established, we decided to test and adapt the setup (robotic arm plus universal AutoIt control script) to the previously mentioned, significantly different scenarios in the laboratory that are described more in detail in the following.
      A central task in every biological, biochemical or chemical laboratory is the transfer of liquids. Mostly, pipets—usually manual pipets—are utilized to fulfill this task. However, as they become less expensive, electronic pipets are now also widely used by researchers around the world. Instead of moving a piston up and down with the thumb, the researcher aspirates and dispenses liquids only by the press of a button. Such electronic pipets are available on the market for less than 300 € and are able to cover the range from 0.1 to 5000 µL. Of course, the casing of the pipet holder could easily be adapted by 3D printing to the requirements of any other chosen manufacturer.
      As shown in Fig. 2a, the casing includes a simple hinge that opens and closes a lockable lid. An integrated electronic switch is controlled via the interfaces available on the robot arm and controls the pipetting button of the pipet. The robotic arm including the pipette was mounted on the linear rail system (commercially available by Dobot) to extend the maximum range of the robot. The system was controlled via the universal AutoIt control script and an additional adapted script. In the completely automated experimental setup, different amounts of a lissamine green dilution in deionized water were transferred from several source tubes (1.5 mL Eppendorf tubes, Eppendorf SE, Hamburg, Germany) to different wells of a 96-well target plate. Two pipetting methods were tested: a classical single pipetting setup, in which a certain volume was aspirated and completely dispensed, and a multipipetting setup, in which a larger volume was aspirated and dispensed in several, smaller portions. The pipetting setup (single or multi-pipetting) can be selected directly via GUIs that are based on the additional AutoIt script. Further settings can be made via additional follow-up GUIs, for example the number of columns and rows of the 96-well plate, number of source tubes or washing steps after pipetting. Since the pipet holder does not yet provide the tip ejection at this stage, the tip was automatically rinsed with deionized water in between each of the different pipetting steps in the wash station.
      Fig. 2
      Fig. 2Experimental setup of the automated application using the desktop robotic arm Dobot Magician to replace employees standardized workflows. Schematic representation (top) and photograph (bottom) of three different application systems (a-c). a) For automated pipetting, conventional electronic pipets (BRAND) are inserted into a pipetting device, which is connected to the robot (bottom). Liquids were transferred from the source tubes into the target plate. The pipet tip was rinsed in the wash station between the pipetting steps (top). The linear rail increases the action radius of the robot. b) To apply samples automatically to the atomic absorption spectroscopy instrument (AAS), the sampling capillary was assembled to the Dobot Magician (bottom) and the sample rack was placed in front of the robot. The sampling capillary was rinsed in the washing station after each measurement (bottom and top). c) Automatic inoculation of bacteria was performed in a laminar flow bench. The inoculation loop installed to the Dobot Magician was heat sterilized by a gas burner and transferred bacteria from source plate to a maximum of two target plates. Source and target plates are identified by a barcode read by a webcam installed to the robotic arm. The suction cup end effector removes and replaces the plate covers.
      To compare the accuracy and precision of the automated pipetting, a highly trained researcher performed the same experiment manually, using the same electronic pipets. In the manual experiment, pipet tips were changed instead of washed between each pipetting step. To quantify the accuracy and precision, the automated and manually transferred liquids in the 96-well plate were measured by an absorbance readout in a suitable plate reader. Fig. 3a shows the result of this comparison: no significant differences as tested by Mann-Whitney U test were observed between the automated and manual setup.
      Fig. 3
      Fig. 3Comparison of results generated by automated and manual processes. No significant differences were found in experiments performed automatically by Dobot Magician compared to manual processes. a) 20 µL of 0.048 mg/mL (1) and 0.06 mg/mL (2) lissamine dilution were dispensed in steps (multi-dispensing) or 100 µL of 0.003 mg/µL (3), 0.006 mg/mL (4), 0.012 mg/mL (5) and 0.048 mg/mL (6) lissamine dilution were pipetted (single pipetting) in a 96-well plate by Dobot Magician using the pipetting device or manually by an employee. Absorbance was measured at 635 nm. b) Standards (2.5 mg/mL (1), 7,5 mg/mL (2), n=8) and samples of cherry juice (Samples) (n=12) were injected into the AAS via a sampling capillary controlled automatically or manually and concentrations of Ca2+ and Mg2+ were recorded. c) E.coli (n=12) and negative controls (NC, 1:1: 80% glycerol, culture medium, n=12) were inoculated from a source plate into 12-well target plates, incubated overnight and bacterial growing was determined by cumulative absorbance at 600 nm.
      The average values and the standard deviations were comparable. We conclude, that both accuracy and precision are comparable for the automated versus the manual setup. No errors could be detected during the pipetting process. We also infer that rinsing the tip, as opposed to exchanging it, does not have a negative effect. If more sensitive substances or experiments are to be performed with this automation system, the setup of multiple rinsing vessels (as in the AAS autosampler) is feasible. Obviously, a more sophisticated automated pipetting option that, for example, facilitated the automated change of pipet tips, would be desirable to eliminate concerns about contamination as far as possible. As a next step in the development, this could be realized by integrating an additional electronic switch into the existing setup.
      Next, we investigated the usability of our setup to act as an autosampler for an AAS instrument. In manual routine measurement, AAS measurements of several samples can be a time-consuming and monotonous task. During the measurement, the sampling capillary needs to be manually placed into the sample container, several buttons need to be clicked in the AAS control software, the AAS instrument performs the measurement for about 2-3 min, the sampling capillary needs to be cleaned and the process starts over for every additional sample. For the automated setup, we placed a sample rack next to the robotic arm (holding up to 24 samples in 50 mL Falcon tubes, Greiner Bio-One GmbH, Frickenhausen, Germany) as well as a wash station to clean the sampling capillary (Fig. 2b). AutoIt controlled both the movement of the robotic arm and the AAS control software SpectrAA. As before, AutoIt furthermore provided the user interface via GUIs. Detailed user instructions via GUIs for example included instructions in which order samples have to be placed in the sample rack, how the robotic arm needs to be prepared, safety instructions, amongst others. Of course, it is also possible for the user to choose the number of samples to be measured from a dialogue box. Instructions how to start up and shut down the system were also included.
      We performed the validation experiment as described in the Materials and Methods section to determine the precision and accuracy. No significant differences as tested by the Wilcoxon test were observed between the automated and the manual setup. Both the average values and the standard deviations did not significantly differ. We conclude, that both accuracy and precision are comparable for the automated versus the manual setup, just as found for the automated pipetting approach. During the whole automated AAS autosampler experiments, no errors in the automated process were overserved. In detail, neither the obtained values nor the visual control during the validation experiment gave any hint of errors. It is thus promising to see, that low-cost devices can be successfully used as autosamplers—as demonstrated before in similar setups [
      • Barthels F.
      • Berthels U.
      • Schwickert M.
      • Schirmeister T.
      FINDUS: an open-source 3D printable liquid-handling workstation for laboratory automation in life sciences.
      ].
      Finally, we tested our system in a more complex laboratory environment. In microbiology laboratories, it is a common task to inoculate bacterial cultures from one container to several others. Usually, an inoculation loop is used to transfer a relatively defined amount of bacterial culture from a source container to a target container. During the manual process, the researcher transfers both the 12-well source and target plate to a clean bench, removes the lids from all plates, unpacks inoculation loop, places the inoculation loop into the first well of the source plate and transfers the loop to the corresponding target well. Then the inoculation loop is discarded, and the process starts over with a new well and inoculation loop.
      As described before, we used a reusable inoculation loop for the automated process that was heat sterilized after each inoculation step. The suction cup served as an end effector for moving plate lids and the webcam was used for barcode recognition. The experimental setup is shown in Fig. 2c. As usual, AutoIt served as a user interface via GUIs and controlled the robotic arm. According to user inputs via GUIs, the desired number of bacterial cultures can be transferred to one or two target plates. AutoIt also controlled the sensory input of the webcam. In more detail, a freeware was used (bcWebCam) to automatically recognize 2D barcodes []. Before initial incubation, a “source plate barcode” was placed on the source plate lid and a “target plate barcode” was placed on each target plate lid. Furthermore, 2D barcodes were used to identify empty trays of the carrier plate. As one of the first steps in the automated setup, the robotic arm moved across all tray positions to automatically identify the barcodes. Thereby, empty trays were automatically recognized as such, in case the user wanted to inoculate only one target plate. In addition, the correct inoculation sequence (first source plate, then target plates) was also automatically identified. In the automated setup, the robotic arm removed the lids of the source plate and the first target plate and inoculated the desired number of wells. Then the lid of the first target plate was closed and the lid of the second target plate was opened and inoculated.
      We inoculated four target plates each manually and automatically in two rounds to judge precision and accuracy. The result showed no significant differences in both average values and standard deviation (Fig. 3c) as tested by non-parametric Levene tests. We conclude that both accuracy and precision are comparable for the automated versus the manual setup. Bacterial growth were found in wells of two negative controls and can be attributed to contamination in the source plate because they were found in the same wells of both the manual and automated target plates. We excluded these outliners when assessing the results.
      Summarizing the three tested automation projects, we can conclude that our system is, in principle, suitable for the automation of fundamentally very different laboratory scenarios. The familiarization of untrained personnel with the fully adapted system proved to be largely unproblematic. While our setup proved to be flexible and user-friendly enough to be used in different laboratory setups, certain drawbacks need to be mentioned.
      First, it must be mentioned that the current system is limited in terms of throughput and speed. Its use is thus limited to applications that are neither in the high-throughput application area nor in time-critical laboratory processes. To give an insight into the time frames needed for the three applications, the automated processes take more time than the manual processes, by the following factors: pipetting process: 3.5, AAS autosampler: 1.6, inoculation of bacteria: 3.0.
      However, our system saves time because laboratory personnel can devote time to other activities while the automated process is running. For example, the manual AAS measurement of 24 samples takes approximately 75 min – time that can be used to work on other processes while the automated setup is running.
      As we knew before, a 4-axis robot is limited in its movements. With additional axes, more samples could be handled in the AAS project. In the pipetting project, more axes would allow a variation of the angle when pipetting.
      In some of the projects, but most specifically in the inoculation project, we also found that the repeatability of the Dobot Magician system was severely limited. This was probably due to the extension of the robot axis by the inoculation loop. Therefore, we could not use 24-well or 48-well plates due to the smaller well size. It may be possible to process a larger plate format using a similar robotic system that has a more accurate spatial repeatability. The model Dobot MG400 could be a conceivable system. It has higher repeatability than the Dobot Magician (0.05 mm versus 0.2 mm), but it costs almost twice as much (approximately 3K € versus 1.5K €) []. Even though this is still in the low price sector, this would increase the price of automation, while simultaneously offering more possibilities.
      We would also like to emphasize that the robotic system used is not a collaborative system. Thus, when using it, care must be taken to ensure that there are no obstacles within the radius of action of the robot arm. In the AutoIt-based dialogue system, this is clearly pointed out to the user. In the future and based on the experience gained here, we will try to address open points such as the missing collaboration, a better repeatability and last, but not least, a further improved usability.
      Despite the limitations mentioned above, we hope that we provided an interesting thought-provoking impulse as to how a system that can be used flexibly in the laboratory can be implemented with extremely few financial resources.
      We also believe that we have been able to develop a well-functioning research system that appears to be well suited not least for teaching purposes.
      Supplemental Files (submit supplemental files and/or videos separately)
      Video: Supplement Rupp et al.mp4

      Declaration of Competing Interest

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      Acknowledgments

      The authors would like to thank the CVUA Sigmaringen for providing measuring time at the AAS and the pleasant working atmosphere and Ms. Eva Lipke and Mr. Tu Truong for their professional contributions to the project. We would also like to thank Mr. Waldemar Rupp and Mr. Stanislav Kukuschkin for the generous support in designing and providing the carrier plate for the inoculation project. The project was funded by a Fit4Research grant provided by the Institute of Applied Research (IAF), Albstadt-Sigmaringen University.

      Appendix. Supplementary materials

      References

        • Thomson Jr., R.B.
        • McElvania E.
        Total laboratory automation: what is gained, what is lost, and who can afford it?.
        Clin Lab Med. 2019; 39: 371-389
        • Lippi G.
        • Da Rin G.
        Advantages and limitations of total laboratory automation: a personal overview.
        Clin Chem Lab Med. 2019; 57: 802-811
      1. Personal communication, Siemens Healthcare.

        • Neubauer J.C.
        • Sébastien I.
        • Germann A.
        • et al.
        Towards a full automation of the elispot assay for safe and parallelized immunomonitoring.
        Methods Mol Biol. 2018; 1808: 237-247
        • Mohammed M.I.
        • Desmulliez M.P.Y
        Lab-on-a-chip based immunosensor principles and technologies for the detection of cardiac biomarkers: a review.
        Lab Chip. 2011; 11: 569-595
        • Wang G.
        • Das C.
        • Ledden B.
        • et al.
        Development of fully automated low-cost immunoassay system for research applications.
        SLAS Technol. 2017; 22: 518-528
      2. https://sila-standard.com/.

        • Bär H.
        • Hochstrasser R.
        • Papenfuß B.
        SiLA: basic standards for rapid integration in laboratory automation.
        SLAS Technol. 2012; 17: 86-95
        • Carvalho M.C.
        Practical Laboratory Automation: Made easy with AutoIt.
        WILEY-VCH, 2017 (ISBN: 978-3-527-34158-0)
      3. https://opcfoundation.org/markets-collaboration/lads/.

        • Carvalho M.C.
        • Eyre D.E.
        A low cost, easy to build, portable, and universal autosampler for liquids.
        Methods Oceanogr. 2013; 8: 23-32
        • Reed C.E.
        • Fournier J.
        • Vamvoukas N.
        • et al.
        Automated preparation of ms-sensitive fluorescently labeled N-glycans with a commercial pipetting robot.
        SLAS Technol. 2018; 23: 550-559
        • Zucchelli P.
        • Horak G.
        • Skinner N.
        Highly Versatile Cloud-Based Automation Solution for the Remote Design and Execution of Experiment Protocols during the COVID-19 Pandemic. 2021; 26: 127-139
        • Councill E.E.A.W.
        • Axtell N.B.
        • Truong T.
        • et al.
        Adapting a low-cost and open-source commercial pipetting robot for nanoliter liquid handling.
        SLAS Technol. 2020; (Nov 192472630320973591)
        • Storch M.
        • Haines M.C.
        • Baldwin G.S.
        DNA-BOT: a low-cost, automated DNA assembly platform for synthetic biology.
        Synth Biol. 2020; 5 (ysaa010)
        • Eggert S.
        • Mieszczanek P.
        • Meinert C.
        • Hutmacher D.W.
        OpenWorkstation: A modular open-source technology for automated in vitro workflows. 8. HardwareX, 2020 (VolumeOctober 2020, e00152)
        • Steffens S.
        • Nüßer L.
        • Seiler T.B.
        • et al.
        A versatile and low-cost open source pipetting robot for automation of toxicological and ecotoxicological bioassays.
        PLoS One. 2017; 12e0179636
        • Chan K.
        • Koen M.
        • Hardick J.
        • et al.
        Low-cost 3D printers enable high-quality and automated sample preparation and molecular detection.
        PLoS One. 2016; 11e0158502
        • Dixit C.K.
        • Kadimisetty K.
        • Otieno B.A.
        • et al.
        Electrochemistry-based approaches to low cost, high sensitivity, automated, multiplexed protein immunoassays for cancer diagnostics.
        Analyst. 2016; 141: 536-547
      4. https://www.dobot.cc/dobot-magician/product-overview.html.

        • Ciulu M.D.
        • Stoicu-Tivadar L.
        • Benis A.
        Automating processes in laboratories with the support of dobot magician.
        Appl Med Inform. 2021; 43 (Cluj-Napoca Bd): 22
        • Zhang C.
        • Wijen B.
        • Pearce J.M.
        Open-source 3-D platform for low-cost scientific instrument ecosystem.
        SLAS Technol. 2016; 21: 517-525
        • Barthels F.
        • Berthels U.
        • Schwickert M.
        • Schirmeister T.
        FINDUS: an open-source 3D printable liquid-handling workstation for laboratory automation in life sciences.
        SLAS Technol. 2019; 25: 190-199
      5. https://www.autoitscript.com/site/autoit-tools/.

        • Carvalho M.C.
        • Murray R.H.
        Osmar, the open-source microsyringe autosampler.
        HardwareX. 2018; 3: 10-38
        • Fleischer H.
        • Quang Do V.
        • Thurow K.
        Online measurement system in reaction monitoring for determination of structural and elemental composition using mass spectrometry.
        SLAS Technol. 2018; (Jan 19, 2472630318813838)
        • Carvalho M.C.
        Miau, a microbalance autosampler. 10. HardwareX, 2021: e00215
        • Carvalho M.
        Integration of analytical instruments with computer scripting.
        SLAS Technol. 2013; 18: 328-333
      6. https://www.beuth.de/de/technische-regel/bvl-l-31-00-10/1251196.

      7. http://www.bcwebcam.de/en/index.html.

      8. https://variobotic.de/industrieloesungen/dobot-mg400/.