Building a biofoundry

Abstract

A biofoundry provides automation and analytics infrastructure to support the engineering of biological systems. It allows scientists to perform synthetic biology and aligned experimentation on a high-throughput scale, massively increasing the solution space that can be examined for any given problem or question. However, establishing a biofoundry is a challenging undertaking, with numerous technical and operational considerations that must be addressed. Using collated learnings, here we outline several considerations that should be addressed prior to and during establishment. These include drivers for establishment, institutional models, funding and revenue models, personnel, hardware and software, data management, interoperability, client engagement and biosecurity issues. The high cost of establishment and operation means that developing a long-term business model for biofoundry sustainability in the context of funding frameworks, actual and potential client base, and costing structure is critical. Moreover, since biofoundries are leading a conceptual shift in experimental design for bioengineering, sustained outreach and engagement with the research community are needed to grow the client base. Recognition of the significant, long-term financial investment required and an understanding of the complexities of operationalization is critical for a sustainable biofoundry venture. To ensure state-of-the-art technology is integrated into planning, extensive engagement with existing facilities and community groups, such as the Global Biofoundries Alliance, is recommended.

1. Introduction

Currently, over 40 countries have national strategies relating to the ‘bioeconomy’ (the economic potential of bioscience) and/or synthetic biology, including the USA (12), the UK (3) and the Australian Council of Learned Academies (ACOLA) report (4). In many of these strategies, the growth of synthetic biology capabilities is identified as critical to scientific and economic competitiveness. A comprehensive, national infrastructure platform is considered by most to be essential to developing and growing synthetic biology capacity. The high-throughput capability afforded by access to a biofoundry can satisfy this goal.

A biofoundry is an integrated molecular biology facility that includes robotic liquid-handling equipment, high-throughput analytical equipment, and the software, personnel and data management systems required to run the equipment and broader biofoundry capabilities. Biofoundries marry synthetic biology with automation engineering to create new high-throughput biological solutions that help to build and strengthen a Design-Build-Test-Learn (DBTL) approach to biological engineering (Figure 1).Figure 1.The Design-Build-Learn-Test cycle. Biofoundries typically focus on the BUILD and TEST aspects of the cycle, but this is not a firm rule; a facility might, for example, focus more on LEARN capabilities by developing data analysis and machine learning methods.Open in new tabDownload slide

The Design-Build-Learn-Test cycle. Biofoundries typically focus on the BUILD and TEST aspects of the cycle, but this is not a firm rule; a facility might, for example, focus more on LEARN capabilities by developing data analysis and machine learning methods.

The Design-Build-Learn-Test cycle. Biofoundries typically focus on the BUILD and TEST aspects of the cycle, but this is not a firm rule; a facility might, for example, focus more on LEARN capabilities by developing data analysis and machine learning methods.

Biofoundries are gaining popularity around the world, with academic and commercial facilities being established across North America, Europe and Asia-Pacific regions. A Global Biofoundries Alliance (GBA) (5) for noncommercial biofoundries was launched in 2019, with 16 founding members (67); and has already grown to 27 members in 2020. Core aims of the GBA include developing best practices across facilities, sharing of information and resources and enhancing visibility and support for these facilities.

Establishing a biofoundry is a significant investment and requires more than simply setting up a well-equipped physical space. The emphasis on high-throughput methods requires concomitant attention to software, protocols and the integration of physical and digital infrastructures to efficiently prepare and track samples. In this respect, biofoundries are at the forefront of a paradigm shift in biological engineering toward a more automated, design-focused venture.

In this review, we offer an overview of key technical, organizational and operational issues relating to the setup and running of a biofoundry. Our focus is on academic rather than commercial biofoundries, although there is clearly overlap between the two, and at least some of our recommendations might apply to commercial foundries. We distinguish between academic and commercial biofoundries primarily with respect to their locations, with academic foundries being located in academic institutions or government laboratories, and commercial facilities operating outside of an academic context. There are broad differences in the target client base, funding and profit models, and sustainability/growth objectives of academic and commercial biofoundries. Academic biofoundries typically focus on supporting the research community and translational activities, while commercial biofoundries have more of a focus on commercial clients and investment return. The recent creation of a GBA to specifically promote and support noncommercial biofoundries suggests there are common challenges for the long-term development and sustainability of these facilities. Our goal in this review is to share what we have learned with respect to establishing and running academic biofoundries.

2. Rationale for Establishing a Biofoundry

Biofoundries deliver capabilities that allow for an accelerated approach to synthetic biology research and application. They also facilitate development of economically important bioengineered products and organisms. This underpins the strategic and economic drivers for acquiring this capability, and there is consequently growing interest and activity worldwide in their establishment. However, a biofoundry requires significant—and ongoing—investment, and so it is important to establish a clear justification and a solid business case before proceeding. As part of business case development, the following key issues should be considered:

2.1 Anticipated throughput (market demand)

The equipment found in most biofoundries enables a significant scale-up in throughput. The throughput capacity of biofoundries is rarely fully utilized; understanding the need or desire (market demand) is critical in order to ensure long-term, efficient use of the capacity of a biofoundry. This capacity typically exceeds the needs of a single laboratory, so identifying collaborators and/or clients, internal and external to one’s department or institution, is an important step to determining the viability of a facility. Market saturation, too many biofoundries for a given research community or insufficient client availability in smaller research communities, could result in facility failure and loss of the significant establishment investment. One way to avoid such an outcome would be a coordinated funding scheme by national granting agencies, preferably as part of their national bioeconomy strategy.

2.2 Scale of investment

Biofoundries require significant time and human capital to set up and operate. The quantum of funding required to establish and support a biofoundry is contingent on the anticipated scale and reach of the facility, but necessarily extends beyond the purchase of robotic, high-throughput equipment to include consumables, software and support for skilled personnel to set up and run the facility. When seeking investment, provisions need to be made for each of these elements. Long-term support from one’s home institution, including an understanding of the nature and scale of the venture, is critical.

2.3 Nature of experiments

Through the use of automation to perform high numbers of repetitive, standardized tasks, biofoundries can dramatically increase the throughput and design space for biological engineering (6–8). Modularization of workflows allows for a mix and match approach and introduces flexibility in services offered in the DBTL engineering biology cycle (Figure 1). However, some types of experiments are more amenable to high-throughput, automated workflows than others. Research programs that require continual small (or large) adjustments to experimental workflows should consider whether a full biofoundry is the most appropriate solution to their needs, or whether bringing key, individual pieces of high-throughput equipment, such as liquid handlers, into their workflows might satisfy their requirements. Focusing on bringing in key, individual pieces of equipment can also be a strategy for lower-resourced settings looking to grow their biological engineering capabilities.

3. Institutional and Funding Models

Biofoundries have often been started with a block of strategic funding from a public sector entity for the purchase of key robotic equipment, often with insufficient reference to long-term sustainability. The initial block funding must be matched with other sustained sources of funding, because the costs associated with running a biofoundry are high: they require specialized staff, expensive equipment and large volumes of consumables, as well as longer-term equipment maintenance and upgrade costs. These costs have been singled out as a challenge for ensuring sustainability (7).

Approaches vary for sourcing funding, and hinge to some extent on national and institutional funding structures and priorities. For biofoundries based in individual research groups or academic institutions, ongoing funding may be part of key infrastructure allocations. Activities may be heavily subsidizd, or grant-funded, to ensure access by institute members. Biofoundries with a national outlook may be funded by combinations of public/government funding and grant funding, with the key driver to support national capability across academia and industry.

For biofoundries without these ongoing support mechanisms, the key to medium- and long-term sustainability is the creation of a core client base and an appropriate service model (see Section 4 Development Strategies and Client Engagement). Appraisal of the client base should extend beyond the host institute, to the broader national and international research communities. These strategies require significant client/community engagement. They can also lead to identification of additional funding sources and alternative funding models.

4. Development Strategies and Client Engagement

Central to the design of a biofoundry is establishing the mode(s) of engagement and interaction with clients and collaborators, be they academic and/or commercial users. In developing service models, it is critical to bear in mind that effective engineering involves a clear understanding of user needs/wishes (i.e. user-centered design). This said, the cutting-edge nature of biofoundries means there may still be limited understanding within the broader research community of what a biofoundry could do for them (see Section 11). Many researchers are unaware that the technology allowing for automated execution of key genetic engineering laboratory protocols is already available and can be routinely employed. Ongoing discussion with the current and prospective user base and an iterative development strategy is required. As the market scale and client needs often require parallel development to the physical infrastructure, it is critical to start small and grow organically with a biofoundry facility so that the facility meets the need without overcapitalization.

4.1 Services available

The focus of a biofoundry can be narrow or expansive. In practice, different biofoundries tend to specialize in different facets of synthetic biology, focusing, for example, on a specific organism or application area, or on different types of services or foundational technologies within the biodesign space (9). The specific niche of a biofoundry should be determined by client need and available equipment.

Much of the core work of current foundries focuses on ‘build’ and ‘test’ capabilities, specifically high-throughput construction of DNA componentry, transformation of cell lines and basic product validation. Facility clients often want to do their own analysis (‘test’) work, or want minimal in-house testing done by the biofoundry. The modularity of synthetic biology workflows can allow clients to pick and choose when and where to enter and leave a service pipeline, and to build custom pipelines from a set of interchangeable modules (Figure 2).Figure 2.Modular nature of biofoundry services. Services offered by a given biofoundry can be easily grouped into functional modules and clients may be allowed to mix and match modules and/or services to fit their needs better.Open in new tabDownload slide

Modular nature of biofoundry services. Services offered by a given biofoundry can be easily grouped into functional modules and clients may be allowed to mix and match modules and/or services to fit their needs better.

Modular nature of biofoundry services. Services offered by a given biofoundry can be easily grouped into functional modules and clients may be allowed to mix and match modules and/or services to fit their needs better.

An example of a basic high-throughput workflow could include DNA synthesis, construct assembly, bacterial/yeast transformation, colony picking, classification and DNA/RNA recovery. Biofoundries with limited equipment may choose to partner with, or outsource to, other facilities to complement or extend workflows. This could include sequencing, cell sorting, proteomic or metabolomic studies and data analysis. Some of the most-requested testing capabilities are flow cytometry, microplate reading assays, fluorescence-activated cell sorting and high-throughput micro-bioreactors. If a given facility has a specific focus or application area (e.g. strain construction for chemical production; large construct building for plant transformation), then the available equipment and workflows developed should support those applications. As with other aspects of biofoundry development, the analytical capabilities attached to a given facility should align with client needs.

4.2 Service models

Biofoundry service models are similar to other scientific service delivery platforms and core facilities, in that the model depends on the funding and the client base. Models in use to date offer differing degrees of access to the facility, ranging from equipment-only access and training programs to full in-house service. Funding models range from full cost recovery (with or without profit) to partial cost recovery to fully subsidized access. For facilities with both academic and industry clients (10), it is common to see a tiered-cost model combining several different cost-recovery levels. Lower recovery rates are typically geared toward academic users and higher recovery from industrial/commercial clients, using profits to subsidize the academic clients.

4.3 User engagement

Client engagement and relationship management are central to any service model. Early discussion points in project engagement include what are client ‘must-haves’, like-to-haves’ and what is not necessary with respect to project outcomes. At a higher level, current client needs, and projected future client needs should be drivers for longer-term planning and growth. This includes plans for new equipment acquisition, protocol development and staff recruitment.

For facilities operating in a cost-recovery model, client engagement necessarily includes discussions around the cost per sample/plate/unit. As automation-oriented consumables can be more costly per unit, it can be useful to provide comparative costing (both time and labor) for the same protocols to be conducted without automation. Wherever partial costs are absorbed by core biofoundry funding, the client should be made aware of full project cost relative to charged cost. Furthermore, it is imperative to engage clients in discussions about forward planning for biofoundry access, e.g. in grant applications. The necessity of planning for associated costs should not be underestimated by the service provider or the client. It should be regarded as similar to planning access to other core facilities or services with such as animal houses or microscopy suites.

4.4 Early collaborators

In the development of core biofoundry protocols, it can be useful to establish a small number of early collaborations that enable the development of key workflows and standard operating procedures. This period can be used to determine approximate failure rates in processes, to understand the cost and time commitments of different levels of troubleshooting, and to develop quality control measures. This information can be crucial for managing future client relationships and allowing for realistic projections on workflow time frames and associated costs.

5. Site Considerations

5.1 Centralized versus distributed model

Biofoundries can be physically centralized or distributed. Most facilities are currently centralized, that is, all staff and machines are in one location/building/room. Most managers see this as an advantage with respect to managing communication between researchers, simplifying workflows. On the other hand, there are examples of successful distributed biofoundries, sometimes with significant geographical distribution. Advantages of a distributed model are that it can leverage established capability at distinct sites and service a broader cross-section of the research community. However, governance issues can arise with distributed models if multiple partner organizations have differing governance approaches or processes. Furthermore, having sites in different geographical locations may mean having to comply with multiple sets of regulations. In either model, often the biofoundry facilities are colocated with or housed within a research institute, or with a local industry. This embedding can facilitate access by users/clients and can promote collaborative information exchange. A further benefit is that it can also lend the biofoundry the credibility of the host organization. This is particularly advantageous in the earliest days of operation and establishing a client base.

5.2 Physical requirements

A second site consideration for biofoundries is the physical attributes of the space housing the facility. While a standard molecular biology lab can be turned into a biofoundry, depending on resource availability we recommend choosing premises with following features: (i) containment infrastructure necessary for the microorganism Risk Group (1112) along with institutional and regulatory authority approval to work with genetically modified organisms (GMOs); (ii) at least 20 benches worth of space (for immediate development); (iii) climate control/air conditioning; (iv) vacuum and compressed air lines; (v) access to three-phase power supply; (vi) sufficient Ethernet ports for each instrument to be networked; (vii) enough space for at least 1.8 m worth of biosafety cabinet (BSC) class 2 space; and (viii) as open a layout as possible.

6. Personnel Considerations

Although it might not seem intuitive for a facility focused on automation, personnel are critical to the success of a biofoundry. Indeed, it could be argued that staff are a more important investment than automated equipment in order to build knowhow and ensure continuity in operation over time. There has been a tendency to overlook this in biofoundry establishment, requiring post-hoc securing of large and long-term funding.

6.1 Key roles

On average, a small biofoundry will likely require a team of 5–10 people, with larger facilities and more distributed foundries involving many more. Typical roles in a biofoundry include a high-level manager to oversee biofoundry operations and assist with business development, automation specialists with experience in lab robotics, a software engineer with expertise in Laboratory Information Management Systems (LIMS) and data management, data scientists for managing data and integrating ‘learn’ capabilities, technicians to perform experiments, potentially a system integrator to integrate hardware and software (given that ‘standardized’ biofoundry solutions do not currently exist) and a scientific director (often appointed as a part-time position). The exact distribution among managers, researchers/developers and technicians will vary across facilities, but with the current challenges facing biofoundry setup it is not unusual to see a focus on employing researchers/developers. This distribution may change as the focus of a facility shifts from development to day-to-day operations. The development strategy of a given biofoundry (see above) will likely necessitate long-term retention of automation, software and data specialists alongside technicians. Ongoing training and staff development pathways are also factors to consider in the strategy.

6.2 Knowhow and tacit knowledge

While the prospect of automated biofoundries might conjure up a future requiring little human input, any such future is a long way off at best. The embodied knowhow that staff bring to running a biofoundry is far from trivial. This knowhow is manifest in several ways, from understanding how the particular goals and biological idiosyncrasies of a project might influence workflow design, to being able to disentangle biological and mechanical sources of failure when troubleshooting, to orchestrating all the work needed behind-the-scenes to manage what at a distance might seem like ‘seamless’ workflows (13). In his doctoral study of automation in a UK biofoundry, Chris Mellingwood highlighted the ‘amphibious’ skills of foundry operators, needing to have deep understandings of both ‘wet’ biological and ‘dry’ robot behaviors (14). With the new and complex configurations of physical, digital and biological elements being developed in biofoundries, the ability to straddle ‘wet’ and ‘dry’ domains is a rare and critical form of expertise.

6.3 Job security and career prospects

Ensuring job security for biofoundry staff is critical to the smooth operation of a biofoundry, not least because of the significant knowhow they build up over time (see above). However, retaining qualified staff can be challenging, not just from a financial perspective but also in terms of supporting career aspirations. Building and running a biofoundry requires considerable scientific and technical talent, but career paths for staff in academic biofoundries are not yet well established. Working across multiple projects in a service role can make it challenging for a young scientist/engineer to carve out the kind of specific identity or publication output typically associated with a career in academia. (See Hilgartner (15) for a discussion of similar challenges faced during the Human Genome Project). This said, there continue to be opportunities to contribute to fundamental knowledge generation around the setup and operation of biofoundries as well as for addressing important scientific questions, as evidenced by several papers in this collection (OUP Synthetic Biology; Biofoundry special issue). Recommended steps for biofoundry managers to pursue include developing workforce retention plans to ensure continued service delivery, working with individuals to establish career development plans, striking an appropriate balance between their service and knowledge generation activities, and promoting opportunities to contribute to research articles.

7. Hardware Considerations

Although it might not seem immediately obvious, the central piece of equipment in a biofoundry is arguably the microplate. Rather than microcentrifuge tubes, which are the core vessels in a classical molecular biology lab, standardized SBS footprint microplates (Standards ANSI/SLAS 1-2004 through ANSI/SLAS 4-2004), with 96, 384 or 1536 wells per plate, are the primary vehicle for samples in a biofoundry. Thus, machines and software under consideration for a biofoundry must be microplate compatible. This orientation around the microplate is both a technical consideration and a major shift in experimental design thinking (see Section 11).

7.1 Liquid handlers

The first major hardware investment for a prospective biofoundry is likely to be in liquid-handling robots. Most procedures performed with manual pipettes can be adapted to automated pipetting or acoustic liquid handlers. There are numerous instrument options at a wide range of price points; generally speaking, robot size and reliability are a function of price (with diminishing returns). There is a growing push to diversify the development and availability of Open hardware and accompanying low-cost automation platforms, which may offer opportunities to develop lower-cost biofoundries (916–19).

The types of protocols being planned for the facility should factor into purchasing decisions for liquid-handling robots. For instance, if a wide range of protocols are required, then a handler with a fair degree of adaptability will be required. Certainly, this may be the case in the early stages of biofoundry development where a single pipetting handler may be used to execute most processes. Depending on the instrument, the handler adaptability may be conferred by deck accessories or interchangeable parts. Or adaptability may be conferred in the range of motion and pipetting head selectivity. For instance, a handler may have interchangeable parts on the pipetting head for selection of pipette tips by number of tips or by tip size (pipetting volume). Alternatively, a handler may have a single pipetting head that can vary the number and size of tips selected. As part of this, consider the throughput expected on the instrument. Will it be necessary to use a 96-barrel pipetting head or will lower (8 or 12-barrel) or higher (348-barrel) pipetting capacity be optimal? These considerations should be tempered with research into the types and expense of consumables required for the instrument and commercial kits to be used in processes.

7.2 Additional equipment

Treating liquid handlers as the central element of a given workflow, attention can then turn to additional high-throughput equipment needed to realize specific elements of the DBTL cycle. For example, a common biofoundry protocol might be the construction of genetically modified Escherichia coli in 96-well plates. Each step of this process can be mapped and associated with specific pieces of automated hardware. The input parts are typically either synthetic DNA with minimal pre-processing required or DNA amplified by PCR from a source template using a microplate-compatible thermal cycler. A liquid handler with cherry-picking capabilities can then be used to combine specific DNA parts with a plasmid backbone, followed by a round in the thermal cycler to ligate the fragments. Organism transformation and plating can then be done using a liquid handler with appropriate heating and shaking capabilities. For colony picking, there are automated colony picking robots (Figure 3). Repeating this process mapping exercise for each of the planned core workflows can help to prioritize equipment purchases. It is also important to consider what basic analytical equipment might be required for initial assessment of the generated outputs. Some biofoundries make lists available of the specific equipment they use (1020–22).Figure 3.Mapping of a manual workflow onto an automated one. This figure depicts a typical mapping exercise where a DNA assembly and subsequent bacterial transformation is being considered for automation. In the first step, the reagents used in tubes are mapped to a microplate layout. Next, instead of manual pipetting automated dispensing is implemented. For the small volumes of a PCR reaction or DNA assembly, an acoustic liquid handler would be good choice. Next, a thermal cycler with 96-well or 384-well plate format is required. For workflows that include DNA amplification by PCR, a quality control step to assess amplified fragments is included. In low-throughput, manual methods this would be done using standard gel electrophoresis. In an automated setup, this can be done in an automated DNA analyzer in 96-well format. Next, the heat shock reaction is be setup and executed using pipettes in manual workflow, whereas this can be executed by an automated liquid handler. Finally, after transformation and overnight growth colonies are picked—using your tool of choice out of a petri dish or using an automated colony picker in the automated alternative.Open in new tabDownload slide

Mapping of a manual workflow onto an automated one. This figure depicts a typical mapping exercise where a DNA assembly and subsequent bacterial transformation is being considered for automation. In the first step, the reagents used in tubes are mapped to a microplate layout. Next, instead of manual pipetting automated dispensing is implemented. For the small volumes of a PCR reaction or DNA assembly, an acoustic liquid handler would be good choice. Next, a thermal cycler with 96-well or 384-well plate format is required. For workflows that include DNA amplification by PCR, a quality control step to assess amplified fragments is included. In low-throughput, manual methods this would be done using standard gel electrophoresis. In an automated setup, this can be done in an automated DNA analyzer in 96-well format. Next, the heat shock reaction is be setup and executed using pipettes in manual workflow, whereas this can be executed by an automated liquid handler. Finally, after transformation and overnight growth colonies are picked—using your tool of choice out of a petri dish or using an automated colony picker in the automated alternative.

Mapping of a manual workflow onto an automated one. This figure depicts a typical mapping exercise where a DNA assembly and subsequent bacterial transformation is being considered for automation. In the first step, the reagents used in tubes are mapped to a microplate layout. Next, instead of manual pipetting automated dispensing is implemented. For the small volumes of a PCR reaction or DNA assembly, an acoustic liquid handler would be good choice. Next, a thermal cycler with 96-well or 384-well plate format is required. For workflows that include DNA amplification by PCR, a quality control step to assess amplified fragments is included. In low-throughput, manual methods this would be done using standard gel electrophoresis. In an automated setup, this can be done in an automated DNA analyzer in 96-well format. Next, the heat shock reaction is be setup and executed using pipettes in manual workflow, whereas this can be executed by an automated liquid handler. Finally, after transformation and overnight growth colonies are picked—using your tool of choice out of a petri dish or using an automated colony picker in the automated alternative.

Figure 4.The biofoundry funnel. This diagram outlines how different elements discussed in this guide interact with each other. They have been listed in the two central columns, reflecting early- (left) and mid- to late-stage (right) activities, noting that there is often need to parallelize and prioritize different parts of the funnel to match the developing situation. Open in new tabDownload slide

The biofoundry funnel. This diagram outlines how different elements discussed in this guide interact with each other. They have been listed in the two central columns, reflecting early- (left) and mid- to late-stage (right) activities, noting that there is often need to parallelize and prioritize different parts of the funnel to match the developing situation.

The biofoundry funnel. This diagram outlines how different elements discussed in this guide interact with each other. They have been listed in the two central columns, reflecting early- (left) and mid- to late-stage (right) activities, noting that there is often need to parallelize and prioritize different parts of the funnel to match the developing situation.

All biofoundry equipment should be procured with potential future automation integration in mind. It is also worth remembering that equipment expenses are not restricted to the purchase price of a machine. Consumables can be costly, and in a high-throughput biofoundry, they will be used in greater volumes than in a nonautomated laboratory. Maintenance contracts for machines can also run as high as 15% of the original purchase cost per annum. Some foundries choose not to pay maintenance costs on individual equipment, and instead include potential maintenance costs in annual financial planning.

8. Software Considerations

Software is critical to running a biofoundry. Key software components include (i) biodesign automation software, for different elements of the DBTL cycle, (ii) an LIMS, for sample and workflow tracking, (iii) DNA screening software for biosecurity considerations and (iv) a web portal for client interaction. Ideally, these four systems should be interlinked.

Some biofoundries are opting to design their own customized software, requiring a dedicated, in-house development team. There is an ever-growing list of software tools to assist with different elements of the biodesign automation process (23). There is also widespread sharing within the community, aiming at ever-important standardization. However, there is currently no off-the-shelf software that satisfies all requirements for a typical biofoundry. Furthermore, numerous gaps have been noted in the integration of existing computer-aided design tools (24). Considerable effort is typically required to operationalize any tool (or set of tools) in a facility, especially given that much of the freely available software is ‘buggy’ and not reliably maintained. There are also broader questions being raised about the suitability of key metaphors currently underpinning the development of DNA design tools, in particularly challenging the idea of DNA as linear text (25). Alongside the refinement of existing tools, some foundational reconceptualization of DNA design tools may be needed to achieve ambitions of improved rational design. Before committing to a specific software approach in a new biofoundry, we recommend consulting with existing biofoundries to determine current state-of-the-art and review operational experiences, as these are changing rapidly.

9. Automation and Integration Strategy

Initial funding for and excitement around biofoundries most often relates to the purchase of high-throughput equipment. However, such equipment can be challenging to install and integrate into biofoundry operation and requires staff with considerable experience (9). In the first instance, machines must be set up for reliable use, which includes tasks like figuring out which brand(s) of consumables to use. Any given machine must be integrated with potentially multiple pieces of software (see above). And developing reliable protocols can require fundamental re-design of experiments to account for reaction volume kinetics and to match equipment capacity (2627).

It is also important to consider the level of automation desirable and/or feasible within a given biofoundry. Developing a fully automated engineering platform can be expensive, time-consuming to operationalize, and relatively inflexible when it comes to offering clients access to modular, customizable services. A productive approach can be to start small in terms of both equipment and automation—slowly growing a ‘machine park’ piece-by-piece as needs arise and moving. In tandem with this, an automation strategy can evolve from discrete islands of automation, for processes with clear and repeated workflows, toward more integrated systems (28).

10. Data Access Considerations

A fully operational biofoundry will produce large amounts of data. Robust mechanisms for managing, storing and accessing data thus become critical. Simply understanding how to use and format data produced by a foundry can be a daunting task, particularly if wanting to ensure they conform to FAIR principles (findability, accessibility, interoperability and reusability) (29). Developing databases, data management tools and governance structures require bioinformatics and data science expertise. With the projected volumes and quality of data being produced, artificial intelligence algorithms are increasingly being promoted to help mine data and improve the ‘Learn’ capabilities of biofoundry design-build-test cycles (30).

10.1 Data inputs

The data required to complete builds will generally be provided to the biofoundry by the client or will be available on publicly accessible genetic sequence databases. If physical DNA samples (and in some jurisdictions, digital sequence data) are provided by the client, then the onus will likely fall on them (not the biofoundry) to ensure that the samples were obtained in accordance with relevant laws, recognizing that countries have sovereign authority over their genetic resources and this requirement can be put into service agreements or contracts signed with the clients (Convention on Biological Diversity (CBD) (31)). These include any import/export rules and domestic genetic resource access and benefit-sharing policies that have been implemented under the UN’s CBD and its supplementary Nagoya Protocol (32).

Multiple UN forums are currently discussing capturing ‘digital sequence information’ (including genetic sequence data) in the same access and benefit-sharing regime that regulates access to physical biological samples (33). This could have a major impact on the operation of biofoundries and synthetic biology more generally (34). Some countries are already regulating access to genetic sequences from biological samples originating in their countries (e.g. India, Malaysia and Kenya), the use of which may attract benefit-sharing obligations (35). These data-input considerations will be particularly important for any clients wishing to commercialize downstream products. The client should be able to account for the provenance of all genetic resources and associated genetic sequence data used in the R&D process.

10.2 Data outputs

Ownership of data produced by a biofoundry will largely depend on its service model and legal obligations. If a client engages the biofoundry collaboratively, then intellectual property considerations will likely be negotiated on a case-by-case basis (along with permissions, warranties, assignments, licenses and indemnities) and outlined in an MoU and/or service agreement between the client and the biofoundry prior to engagement.

Data output considerations are also important for academic clients not intending to commercialize their research, especially given synthetic biology community’s tendency toward open data. They will need to consider whether the data produced by the biofoundry will be published on public access genetic sequence databases, noting that there may be restrictions on some genetic sequence data from genetic resources originating in certain countries (35). Publication of genetic sequences may be a prerequisite to publication in academic journals and can also be a requirement of some research funding agencies.

11. Cultural and Training Considerations

11.1 Outreach

Biofoundries are at the forefront of a conceptual shift in bioengineering design. In developing robust, flexible, high-throughput capabilities, the very nature of what can be achieved in a single ‘experiment’ is rapidly expanding. While this might be clear to the operators of biofoundries and their current client base, such a conceptual shift has not yet taken root across the life science disciplines that deal with the engineering (genetic, metabolic, protein and others) community at large (36). To combat this, biofoundries need to engage more proactively, not just in marketing their services, but also in the use of educational tools such as workshops, conferences, lectures and webinars. These can all assist potential clients in determining how a biofoundry may facilitate their research. At a more granular level, direct engagement with potential clients to discuss their research goals and concerns is beneficial. Automation necessarily involves some compromises (26) and understanding what compromises potential clients would be reluctant to make may help a given biofoundry better tailor its services and workflows to prospective users.

Beyond individual biofoundries, the GBA, through a working group, has an outreach program focusing on industry engagement, policy, safety and security, and the public (5). Two of the aims of the program are to provide impact greater than which individual biofoundries can achieve, and to increase the awareness about the role and importance of biofoundries. Furthermore, sharing best practices across biofoundries through forums like the GBA could prove advantageous, as a means of recruiting a broader client base. These efforts are two-way: as well as promoting understanding among potential clients regarding the possibilities enabled by biofoundries, work is undertaken to understand resistance and hesitation to using biofoundries among the broader research community.

11.2 Training

Several academic biofoundries began with the assumption that once their facilities were up and running, there would be widespread demand for their services across (and beyond) their home institution. However, in academia, the design and allocation of projects do not always map neatly onto the enlarged design space made possible through biofoundries. For example, the design space for a protein engineering PhD is frequently cast as ‘one student, one protein’ when biofoundry capabilities can allow for larger and more combinatorial efforts. Including training in the design of research experiments and projects should thus be seen as a core service offered by a biofoundry. This can be achieved in several ways, for example: working with academic researchers to design biofoundry experiments that are suited to PhD or master’s projects, and designing student and postdoctoral residency or internship programs to provide basic biofoundry training. These strategies will provide a base for generational change in approaches to bioengineering.

12. Biosafety and Biosecurity Considerations

12.1 Biosafety

Biosafety is about preventing ‘accidental interactions between dangerous biological agents and other organisms or the environment’ (37) and can refer to biosafety practices within the lab and/or risk mitigation for modified organisms used outside the lab (34). However, it is recognized that risk mitigation is not in of itself sufficient, as risk identification and response to unpredicted consequences of genetic mutation (38) is necessary.

The biosafety risks and associated containment considerations for biofoundries will largely depend on the host organisms with which the biofoundry intends to work, but may also depend on the DNA expression products (e.g. constructs that increase or confer pathogenicity or virulence) (11). The most common host organisms (e.g. lab strains of E. coliSaccharomyces cerevisiae and Bacillus subtilis) are generally considered safe to handle at Biosafety Level 1 (BSL-1). Different jurisdictions will have different standards for working with GMOs and different requirements for biosafety accreditation. In Australia, for example, biocontainment facilities are accredited by the Office of the Gene Technology Regulator (OGTR), which requires an Institutional Biosafety Committee (IBC) to oversee the activities of the facility.

Internationally, the Cartagena Protocol on Biosafety (2000) to the UN’s CBD (31) regulates the safe transfer, handling and use of living modified organisms across national borders. Even countries that are not party to the various protocols of the CBD may have implemented related provisions in their national legislation. For example, Australia is party to the CBD but not the Cartagena Protocol, however, Australia’s Gene Technology Act 2000 (Cth) is considered sufficient to meet the requirements for national implementation of the Cartagena Protocol (39). Biofoundries should conduct induction and refresher training on the international and national rules for safe handling and transfer of modified organisms, and appropriate labeling and documentation when sending or receiving biological resources and GMOs.

It is unlikely that a biofoundry will be directly involved in the environmental release of a genetically modified end-product. However, should the genetic constructs be incorporated in a living or nonliving product destined for environmental release, the biofoundry may wish to consider the design and inclusion of intrinsic biosafety measures, such as DNA signatures, barcodes or watermarks coded into the constructs (40).

12.2 Biosecurity

In the context of biofoundries, biosecurity refers to the measures taken to reduce the risk of materials and information being used for nefarious purposes and the dual-use research of concern (DURC). Basic biosecurity measures should include physical limits on who can access the biofoundry, vetting potential employees, compliance with national biosecurity regulations, and conducting DURC awareness training.

Biosecurity measures have typically focused on controlling access to physical samples of human, plant and animal pathogens (e.g. through import and export controls) (40,). This includes multilateral export control regimes focusing on dual-use items of interest. The Australia Group (41), for instance, is an informal association, of more than 40 countries and the EU, with the objective of harmonizing export controls on animal (including human) and plant pathogens and toxins, genetic elements and GMOs, fulfilling obligations under the Biological and Toxin Weapons Convention (42) and Chemical Weapons Convention (43).

The focus of international and domestic controls has now moved beyond the international movement of physical materials to screening orders for synthetic DNA. Biofoundries are key sites for making DNA constructs and are major consumers of synthetic DNA (44). Most commercial DNA synthesis providers follow voluntary guidelines (45–47) for screening orders of double-stranded DNA (37). Despite the increased costs and potential delay to service delivery times, industry groups chose to adopt screening guidelines early to address fears surrounding the misuse of synthetic DNA and a general lack of government oversight (48). Given the shortcomings of synthetic DNA order screening (e.g. it is voluntary, may not include oligonucleotide orders or dsDNA under 200 bp and it is unclear to whom suspicious order should be reported) (48), client screening is becoming a more valuable biosecurity measure than sequence screening (34).

There is no clear responsibility for biofoundries to assume similar screening activities of potential clients or projects; however, there have been calls for every link in the synthetic biology R&D chain to be involved in biosecurity screening (44). Even without hard rules or voluntary guidelines for screening potential clients, collaborators or projects, biofoundries will need to exercise discretion in determining what services they are prepared to provide and to whom (e.g. are they willing to service DIY Bio practitioners?) to mitigate potential liabilities, especially when working with pathogens or pathogen-derived sequences. Biofoundries may determine their own broad set of standards from the outset or choose to make such decisions on a case-by-case basis. Given the speed of technological advancement and the inability of law and policy to keep up, there are many legal ambiguities in this space which require further characterization and research (see e.g. Box 1). Biofoundries should engage with their local authorities to discuss how best to address such ambiguities.

BOX 1

National regulations relating to responsibility for imported biological material—including organisms, cell cultures, and nucleic acids—may affect biofoundry operations. This is particularly relevant where a country does not have a domestic DNA synthesis capability. For example, in Australia, the Federal Government Department of Agriculture biological import permit conditions mandate that the primary importer is responsible for imported materials (https://www.agriculture.gov.au/import). It is unclear just how far downstream this responsibility endures and what such a responsibility would mean for biofoundries, which operate as an intermediary or ‘middle-person’ (making and perhaps testing, but not deploying, engineered constructs/organisms). This regulatory position might make a biofoundry responsible for the activities of clients using organisms and genetic constructs derived from imported biological materials. In the face of ambiguous policies, individual biofoundries may have to negotiate their responsibilities with respect to national policies.

13. Conclusions

Biofoundries are quickly becoming a central element of the synthetic biology landscape. With their focus on automation, they are at the leading edge of a larger reconceptualization and reorganization of bioengineering work that stands to make possible new kinds of knowledge and engineering practice (49). However, building a biofoundry is a complicated process, requiring specialized equipment and expert, dedicated personnel. The running costs can be an order of magnitude higher than a standard laboratory, but biofoundry automation has the potential in principle to increase the speed, reproducibility and reliability of acquired results by several orders of magnitude. An established facility can become the heart of a local or even national synthetic biology program, delivering fast and reliable data for many projects and research teams and forming the basis of new economic development.

In this article, we have outlined numerous considerations and challenges associated with building and operating a biofoundry (Figure 4). As the number of biofoundries continues to grow, the importance of initiatives like the GBA increases. With their focus on sharing best practices, developing standards and business models, and working through legal, security and cultural concerns, such forums can help speed the development and maturation of the synthetic biology ecosystem. When planning a new biofoundry, we strongly recommend engaging with the GBA community and with individual, established biofoundries to gain an understanding of current learnings and best practice. Furthermore, engagement with national roadmaps and broader technological roadmaps (such as the Engineering Biology Research Consortium next-generation bioeconomy roadmap) (2) should assist with planning and developing a sustainable business model that is appropriately integrated within a broader national science and innovation strategy.

Authors’ Contributions

All authors contributed equally to the development, writing and review of this article.

Acknowledgments

We would like to thank Jose Carrasco Lopez, Rosalind le Feuvre, Liz Fletcher, Paul Freemont, Nathan Hillson, Vincent Martin, David McClymont, Filippo Menolascina, Nicola Patron and Hille Tekotte and other members of the Global Biofoundry Alliance for their invaluable insights into inner workings of their respective biofoundries.

Funding

M.B.H. and M.R. acknowledge support from CSIRO’s Synthetic Biology Future Science Platform. E.K.F. acknowledges research support from the European Research Council [EC 616510 ENLIFE].

Conflict of interest statement. None declared.

References

1National Academies of Sciences, Engineering, and Medicine. (2020) Safeguarding the Bioeconomy. The National Academies Press, Washington, DC.

Google ScholarPubMed2Engineering Biology Research Consortium (EBRC). (2019) Engineering Biology: A Research Roadmap for the Next-Generation Bioeconomyhttps://roadmap.ebrc.org (1 July 2020, date last accessed). doi:10.25498/E4159B.3Department for Business, Energy & Industrial Strategy. Growing the Bioeconomy: A National Bioeconomy Strategy to 2030. UK Government. https://www.gov.uk/government/publications/bioeconomy-strategy-2018-to-2030 (1 July 2020, date last accessed).4Gray P. , Meek S. , Griffiths P. , Trapani J. , Small I. , Vickers C. , Waldby C. (2018) Synthetic Biology in Australia: An Outlook to 2030. Report for the Australian Council of Learned Academies (ACOLA). www.acola.org.au (1 July 2020, date last accessed).5Global Biofoundries Alliance (GBA). Global Biofoundries Alliance. https://biofoundries.org/ (1 July 2020, date last accessed).6Hillson N. , Caddick M. , Cai Y. , Carrasco J.A. , Chang M.W. , Curach N.C. , Bell D.J. , Feuvre R.L. , Friedman D.C. , Fu X. et al.  (2019) Author correction: building a global alliance of biofoundries. Nat. Commun., 10, 3132.

Google ScholarCrossrefPubMed7Hillson N. , Caddick M. , Cai Y. , Carrasco J.A. , Chang M.W. , Curach N.C. , Bell D.J. , Le Feuvre R. , Friedman D.C. , Fu X. et al.  (2019) Building a global alliance of biofoundries. Nat. Commun., 10, 2040.

Google ScholarCrossrefPubMed8Freemont P. , Curach N. , Friedman D.C. , Lee S.Y. (2019) These ‘Biofoundries’ Use DNA to Make Natural Products We Need. World Economic Forum. https://www.weforum.org/agenda/2019/10/biofoundries-the-new-factories-for-genetic-products/ (3 December 2020, date last accessed).9Jessop-Fabre M.M. , Sonnenschein N. (2019) Improving reproducibility in synthetic biology. Front. Bioeng. Biotechnol., 7, 18.

Google ScholarCrossrefPubMed10Chambers S. , Kitney R. , Freemont P. (2016) The Foundry: the DNA synthesis and construction Foundry at Imperial College. Biochem. Soc. Trans., 44, 687–688.

Google ScholarCrossrefPubMed11World Health Organization. (2004) Laboratory Biosafety Manual, 3rd edn. Geneva. https://apps.who.int/iris/handle/10665/42981, (1 July 2020, date last accessed).

Google Scholar12Standards Australia. (2010) AS/NZS 2243.3:2010: Safety in Laboratories Microbiological Safety and Containment. SAI Global under Licence from Standards Australia Limited and Standards New Zealand (1 July 2020, date last accessed).13Hammang A. , Frow E. (2020) Mapping Synthetic Biology Workflows: An Experimental Workshop. Arizona State University. https://blogs.sps.ed.ac.uk/engineering-life/files/2020/02/Hammang-Frow-2019-Mapping-Synthetic-Biology-Workflows-circulated-report.pdf (1 July 2020, date last accessed).

Google Scholar14Mellingwood C. (2019) Amphibious researchers: working with laboratory automation in synthetic biology. Doctoral. The University of Edinburgh.

Google Scholar15Hilgartner S. (2013) Constituting large-scale biology: building a regime of governance in the early years of the Human Genome Project. BioSocieties, 8, 397–416.

Google ScholarCrossref16Maia Chagas A. (2018) Haves and have nots must find a better way: the case for open scientific hardware. PLoS Biol., 16, e3000014.

Google ScholarCrossrefPubMed17Gibney E. (2016) ‘Open-hardware’ pioneers push for low-cost lab kit. Nature, 531, 147–148.

Google ScholarCrossrefPubMed18May M. (2019) A DIY approach to automating your lab. Nature, 569, 587–588.

Google ScholarCrossrefPubMed19Storch M. , Haines M.C. , Baldwin G.S. (2020) DNA-BOT: a low-cost, automated DNA assembly platform for synthetic biology. Synth. Biol., 5, ysaa010.

Google Scholar20Johnson J.R. , D’Amore R. , Thain S.C. , Craig T. , McCue H.V. , Hertz-Fowler C. , Hall N. , Hall A.J. (2016) GeneMill: a 21st century platform for innovation. Biochem. Soc. Trans., 44, 681–683.

Google ScholarCrossrefPubMed21Kabisch-Lab. CompuGene Robotic Platform.http://kabisch-lab.de/robotics.html (2 July 2020, date last accessed).22Biofoundry L. What the London Biofoundry Can Do for You.https://www.londonbiofoundry.org/capabilities (2 July 2020, date last accessed).23Appleton E. , Madsen C. , Roehner N. , Densmore D. (2017) Design automation in synthetic biology. Cold Spring Harb. Perspect. Biol., 9, a023978.

Google ScholarCrossrefPubMed24Casini A. , Chang F.Y. , Eluere R. , King A.M. , Young E.M. , Dudley Q.M. , Karim A. , Pratt K. , Bristol C. , Forget A. et al.  (2018) A pressure test to make 10 molecules in 90 days: external evaluation of methods to engineer biology. J. Am. Chem. Soc., 140, 4302–4316.

Google ScholarCrossrefPubMed25Szymanski E. , Scher E. (2019) Models for DNA design tools: the trouble with metaphors is that they don’t go away. ACS Synth. Biol., 8, 2635–2641.

Google ScholarCrossrefPubMed26McClymont D.W. , Freemont P.S. (2017) With all due respect to Maholo, lab automation isn’t anthropomorphic. Nat. Biotechnol., 35, 312–314.

Google ScholarCrossrefPubMed27Meckin R. (2019) Changing infrastructural practices: routine and reproducibility in automated interdisciplinary bioscience. Sci. Technol. Human Values, 45, 1220–1241.

Google ScholarCrossrefPubMed28Alexanian T. (2019) The Case for Modular Lab Automation. Zymergen Technology Team, Medium, 2020. https://medium.com/@ZymergenTechBlog/the-case-for-modular-lab-automation-c34f214e1276 (6 October 2020, date last accessed).29Wilkinson M.D. , Dumontier M. , Aalbersberg I.J. , Appleton G. , Axton M. , Baak A. , Blomberg N. , Boiten J.-W. , da Silva Santos L.B. , Bourne P.E. et al.  (2016) The FAIR guiding principles for scientific data management and stewardship. Sci. Data, 3, 160018.

Google ScholarCrossrefPubMed30Dixon T.A. , Curach N. , Pretorius I.S. (2020) Bio-informational futures: the convergence of artificial intelligence and synthetic biology. EMBO Rep., 21, 1–5.

Google ScholarCrossref31Convention on Biological Diversity (CBD). (1992) Convention on Biological Diversity. Rio de Janeiro, 5 June 1992 Nations, U., Ch_XXVII_8. United Nations Treaty Series 1760. https://treaties.un.org/pages/ViewDetails.aspx? src=TREATY&mtdsg_no=XXVII-8&chapter=27 (1 July 2020, date last accessed).32Nagoya P. (2010) Nagoya Protocol on Access to Genetic Resources and the Fair and Equitable Sharing of Benefits Arising from their Utilization to the Convention on Biological Diversity Nations, U., UNEP/CBD/COP/DEC/X/1 of 29 October 2010, C.N.115.2011.TREATIES-7 of 18 March 2011, Nagoya, Japan. https://treaties.un.org/Pages/ViewDetails.aspx? src=TREATY&mtdsg_no=XXVII-8-b&chapter=27&clang=_en (1 July 2020, date last accessed).33Laird S. , Wynberg R. , Rourke M. , Humphries F. , Muller M.R. , Lawson C. (2020) Rethink the expansion of access and benefit sharing. Science, 367, 1200–1202.

Google ScholarCrossrefPubMed34Gronvall G.K. , Carr P. (2016) Synthetic Biology: Safety, Security, and Promise. CreateSpace Independent Publishing Platform, Scotts Valley, California, US.

Google Scholar35Bagley M. , Karger E. , Ruiz Muller M. , Perron-Welch F. , Siva Thambisetty S. (2020) Fact-finding Study on How Domestic Measures Address Benefit-sharing Arising from Commercial and Non-commercial Use of Digital Sequence Information on Genetic Resources and Address the Use of Digital Sequence Information on Genetic Resources for Research and Development. Annex. Convention on Biological Diversity, Ad Hoc Technical Expert Group on Digital Sequence Information on Genetic Resources,. United Nations, Montreal, Canada, CBD/DSI/AHTEG/2020/1/5. https://www.cbd.int/doc/c/428d/017b/1b0c60b47af50c81a1a34d52/dsi-ahteg-2020-01-05-en.pdf (3 December 2020, date last accessed).36Check Hayden E. (2014) The automated lab. Nature, 516, 131–132.

Google ScholarCrossrefPubMed37Gómez-Tatay L. , Hernández-Andreu J.M. (2019) Biosafety and biosecurity in Synthetic Biology: a review. Crit. Rev. Environ. Sci. Technol., 49, 1587–1621.

Google ScholarCrossref38Ralph E.T. , Guest J.R. , Green J. (1998) Altering the anaerobic transcription factor FNR confers a hemolytic phenotype on Escherichia coli K12. Proc. Natl. Acad. Sci. U S A, 95, 10449–10452.

Google ScholarCrossrefPubMed39Keiper F. , Atanassova A. (2020) Regulation of synthetic biology: developments under the convention on biological diversity and its protocols. Front. Bioeng. Biotechnol., 8, 310.

Google ScholarCrossrefPubMed40Gronvall G.K. (2019) Synthetic biology: biosecurity and biosafety implications. In: Singh S.K. , Kuhn J.H. (eds). Defense against Biological Attacks: Volume I. Springer International Publishing, Cham, pp. 225–232.

Google ScholarCrossref41The Australia Group. Department of Foreign Affairs, Australiahttps://www.dfat.gov.au/publications/minisite/theaustraliagroupnet/site/en/index.html (1 July 2020, date last accessed).42Biological Weapons Convention (BWC). (1975) Convention on the Prohibition of the Development, Production and Stockpiling of Bacteriological (Biological) and Toxin Weapons and on Their Destruction. United Nations Office for Disarmament Affairs, London, Moscow, Washington. http://disarmament.un.org/treaties/t/bwc (3 December 2020, date last accessed).

Google Scholar43CWC. (1997) The Convention on the Prohibition of the Development, Production, Stockpiling and Use of Chemical Weapons and on their Destruction Chemical Weapons Convention. United Nations, Paris. https://www.opcw.org/chemical-weapons-convention (3 December 2020, date last accessed).44Carter S. , DiEuliis D. (2019) Mapping the synthetic biology industry: implications for biosecurity. Health Secur., 17, 403–406.

Google ScholarCrossrefPubMed45International Gene Synthesis Consortium (IGSC). Harmonized Screening Protocol v2.0. Gene Sequence & Customer Screening to Promote Biosecurity.https://genesynthesisconsortium.org/ (1 July 2020, date last accessed).46International Association Synthetic Biology (IASB). The IASB Code of Conduct for Best Practices in Gene Synthesis.http://op.bna.com.s3.amazonaws.com/hl.nsf/r%3FOpen%3Djaqo-7xqpnr (1 July 2020, date last accessed).47Department of Health and Human Services (US HHS). (2010) Screening Framework Guidance for Providers of Synthetic Double-Stranded DNA, Public Health Service Act, 42 U.S.C. 241, Section 301; HSPD-10., 75 FR 62820 pp. 62820–62832. https://www.federalregister.gov/documents/2010/10/13/2010-25728/screening-framework-guidance-for-providers-of-synthetic-double-stranded-dna (1 July 2020, date last accessed).48Bugl H. , Danner J.P. , Molinari R.J. , Mulligan J.T. , Park H.O. , Reichert B. , Roth D.A. , Wagner R. , Budowle B. , Scripp R.M. et al.  (2007) DNA synthesis and biological security. Nat. Biotechnol., 25, 627–629.

Google ScholarCrossrefPubMed49Keating P. , Limoges C. , Cambrosio A. (1999) The automated laboratory. In: Fortun M. , Mendelsohn E. (eds), The Practices of Human Genetics. Springer Netherlands, Dordrecht, pp. 125–142.

Google ScholarCrossref © Crown copyright 2020.This Open Access article contains public sector information licensed under the Open Government Licence v2.0 (http://www.nationalarchives.gov.uk/doc/open-government-licence/version/2/)