Case for Support: A Workflow-based High Throughput Toolkit for Primer Design

Transcription

1 Case for Support: A Workflow-based High Throughput Toolkit for Primer Design Background The Polymerase Chain Reaction (PCR) is one of the core techniques of biotechnology. Software to design primers is required for the efficient application of PCR. Existing primer design software is not flexible enough to meet the ever more complex requirements of novel applications of PCR. We therefore propose to develop a new primer design toolkit that will meet the requirements of 21st century biotechnology. Primer design Primer design is a complex task that requires software tools. The process of primer design involves finding suitable sequences that bind to the correct locations, and also meet a wide range of constraints - for example a pair of primers may have length restrictions, their annealing temperatures should be compatible and within a certain range, they should not make hairpins or primer dimers, they should avoid non-specific binding, their ends should have certain composition, etc. Many publications over the years have made recommendations for good primer design, and have suggested algorithms for the more complex calculations (such as the thermodynamics behind the calculation of melting points, or the sequence similarity searches needed to determine potential non-specific binding). Software tools have been developed, based on these algorithms. There are many existing primer design packages available, ranging from simple tools to help calculate annealing temperatures and other properties of given sequences (such as OligoCalc [6]), to full packages that take the user through the whole process, allow many parameters to be tuned, and produce visualisations of the results. Most packages are desktop applications, often restricted to the Windows platform, and there are surprisingly few programmable packages that can be applied to automated design. Primer3 The most popular non-commercial primer design software is Primer3 [8]. Google Scholar suggests their publication has approx 3,000 citations. Most other freely available primer design software tools are wrappers around Primer3 s core C code. Primer3 s popularity is due to its easy availability, open source code, and extremely parameterisable core. It was first released in 1996 and has been maintained ever since, with the latest release in April 2008 (v1.1.4). The GPL licence ensures that any other code that builds on Primer3 must also be released to the community as open source using the GPL. It is available packaged for Windows, Mac and Linux. The long life span of Primer3 brings advantages and disadvantages. The advantages are years of stable, well tested code, experienced users and a variety of interfaces. The disadvantages are that it has to remain stable and support its user base, and can therefore change only slowly and incrementally. In 2007 modifications to its calculations of melting temperature, salt corrections and effects of divalent cations were published [7]. Version 2.0.0, currently under development, will bring it up to date with C coding conventions, modularise the code, and change the input/output formats to be more automation friendly for programmers. Most of the work on Primer3 since 2000 has been on making web interfaces that allow the vast number of parameters to be sensibly chosen. All these changes to interface formats are not about solving fundamentally new problems. The core of Primer3 remains a surprisingly small amount of code that specialises in application to straightforward PCR reactions. 1

2 Other tools, other problems As biotechnology progresses, and laboratory automation provides new high-throughput capability, then the supporting software needs to advance in step and allow the high throughput workflow needed. Four recent publications demonstrate this trend and show the need for a wider range of primer design tools. Osprey [5] is a primer design system that chose not to use Primer3 in order to make use of more recent models of DNA binding thermodynamics. The authors of Osprey have also developed their own search for unwanted secondary binding, to avoid the heuristic problems of BLAST but still provide an efficient search. Primique [4] is a primer design tool that can also take into account an accompanying set of sequences that the primers should not amplify. Both Osprey and Primique require that the user manually provide sequences, and then primers will be produced that are suitable for amplification of those sequences. BatchPrimer3 and QuantPrime [9, 2] are new tools based on Primer3 which aims to provide more of the complete workflow needed for automation. BatchPrimer allows batch processing and Quant- Prime allows the user to specify transcript identifiers rather than provide the sequences, and filters out primers that would bind to more than one of the transcripts. UniPrime [3] designs universal primers that amplify regions conserved across different taxa, using phylogenetic information. This builds yet more steps into the primer design workflow, such as sequence retrieval from Genbank and sequence alignment with T-coffee. Despite these new software developments, there are an increasing number of advanced scenarios that are not possible using current software. Biologists need assistance with more complex workflows or high throughput automation of primer design. Taqman probes, primers for fusion PCR, chimeric primers and other multistage primer design tasks must still be tackled by hand or with the aid of a computer programmer. Our case study (described later) has 3 PCR stages. The primers at each stage must be consistent with each other and with the constraints, and then be consistent as chimeric composite primers. A search for chimeric primers would need the ability to backtrack and retry, varying parameters. While testing the constraints we want to make sure that we test the cheapest constraints first. We do not want to repeat many time consuming sequence similarity searches to check the primers for initial stage PCRs, only to find that when these are combined for the final fusion PCR the chimeric primers have a hairpin structure. There are also specific hindrances due to individual algorithms. For example, the Primer3 Frequently Asked Questions webpage warns about the difficulties of trying to design long primers (more than 36-mer) with Primer3, due to their methods for calculating melting temperatures. They use a nearest neighbour model which is a good fit for short oligos. Our case study will need much longer primers. Every system will have its own assumptions. What is needed is a modular system that allows new models to be plugged in, and clearly indicates what the assumptions are, how new models can be integrated, and how this will affect the rest of the system. Requirements We therefore need a primer design toolkit that is a flexible and extensible collection of components that can be put together in multiple combinations, coordinated by complex workflows, and available for programmers to use in automated environments. Our proposed toolkit will be open-source. Open source software allows the user confidence in the code, and provides the ability to understand the reasons for any unusual effects when they occur. It also promote the uptake of the tool. The toolkit will make use of and acknowledge existing software such as Primer3 where pos- 2

3 sible, and will contribute back in return to these projects, but will start afresh with the aim of allowing extensibility, modularity and the design of workflows that are currently impossible. At Aberystwyth we have a particular need for such a flexible toolkit, and will demonstrate its success on a case study of high-throughput primer design for seamless gene deletion. Programme and methodology Aims and objectives The four main aims for this proposal are 1. A new flexible primer design toolkit/library that enables the automated design of complex primers (open source, freely available, and providing competition for Primer3, the popular alternative) 2. A case study where this toolkit is applied to automated primer design for seamless gene deletion 3. A complete set of primers designed and published for the seamless deletion of each of the genes in the entire S. cerevisiae genome. 4. A comparison of existing primer design tools with our new toolkit. Methodology Both the principal investigator and the postdoctoral researcher have computer science backgrounds, so this project will be firmly grounded in appropriate software engineering methodology. Specification: We have a wide variety of specification tools available, ranging from the abstract (flow diagrams, UML, workflow languages, process algebras) to concrete (mathematical languages such as Haskell). We would like to specify the tasks, processes, components, interfaces and compositions of a primer design toolkit. We should be able to state pre- and post-conditions for operations, error conditions and result ranges. Research: We are well aware of existing primer design tools. We will carefully examine existing code and publications to be sure we include the knowledge they provide. We will also consult with current users of such applications. Prof Mike Young of the Institute of Biological, Environmental and Rural Sciences is a co-investigator on the project. In addition other people in our own lab and across the University who need to design primers on a regular basis have agreed to both give advice and beta-test the software. Design: The toolkit design will be based on the specification, and will take into account extensibility, modularity, the need for adequate unit testing, the use of open source resources, Web Service standards, platform independent programming languages and the production of automatic documentation. We will also design from the outset for a variety of interfaces to be provided, for full automation and for human use, as a local system or as a remote, Webaccessible application. We will design the toolkit as a library. We would like others to be able to easily isolate parts that they can contribute to and build upon in the future. Implementation: The implementation will follow the design and will involve regular code inspections/walkthroughs, and if necessary, redesign. A version control system will be used and an automated build environment set up. Regular internal releases will ensure the project starts on a professional footing, and provide a trend for continual public development of the toolkit after this initial foundation. Testing: Unit testing will be encouraged. Simple case studies will create more complex test cases. 3

4 Evaluation: We will evaluate the toolkit against the original specification and against Primer3. The Primer3 evaluation will make use of their test suite, and of case studies extracted from the literature. We are aware that there may be differences, due to the algorithms used (eg for melting points or non-specific binding), but these should be explainable. Case study: Seamless gene deletion As part of his PhD project (thesis submitted, Aberystwyth University), Wayne Aubrey has developed a method for the seamless deletion of genes in S. cerevisiae adapted from a previously published procedure [1]. The design of primers for this method cannot be done using existing tools. This technique (without scars that could potentially act as sites for subsequent recombination events or other unwanted side effects), allows multiple genes to be successively deleted. Mr. Aubrey has used his method to make seamless deletion mutants lacking YER152C, YGL202W and YJL060W, and demonstrated its suitability for full automation. This method has been further adapted as outlined in Figure 1 and it is the first method to be described that is amenable to full automation using currently available laboratory automation equipment. Primer design for seamless gene deletion requires two chimeric primers (Figure 1, primers C and D). The precise locations of the upstream (US) and downstream (DS) sequences are flexible, but ideally they should lie immediately adjacent to the extreme 5 and 3 ends of the coding sequence of the gene of interest (GOI). In practise, suitable primers may only be found by permitting some overlap with the coding sequence, in which case it is important that an in-frame deletion is generated at the novel joint between US and DS in Primer C. A 40 bp DS sequence was found by experiment to provide a suitable rate of ura-3 excision in the FOA selection step 5. However, since the DS segment is only 40 bp, the PCR product generated in step 1 must be much longer (ca. 800 bp) to permit the selection of transformants in step 4 (see Figure 1). The tool that we need as part of the process of automating the construction of yeast strains harbouring multiple gene deletions must be designed specifically to address the multi-stage constraints imposed by (a) the seamless deletion procedure and (b) the need for fully automated operation. This tool must work without the need for user intervention. There are no existing packages that are fit for this purpose. Timeliness and novelty Manual primer design is very time consuming and error prone, both in academia and industry, and slows the progress of research. Frequently, the temperatures calculated do not work as planned, and sequence mistakes can cause unusual binding. This wastes expensive equipment time and delays results. Most biologists are still painstakingly using desktop applications such as VectorNTI, or else relying on databases of pre-designed primers. In modern automated laboratories, high-throughput primer design software is becoming essential. Multiplex primer design must take account of many sequences concurrently and chimeric primers must meet many constraints. Primer3 has filled the needs of biologists with simple PCR applications for the past 10 years. There is now a need to expand the range of tools available. With a dedicated library we provide a basis for automation solutions, and a basis for bioinformaticians to take and wrap with human friendly interfaces, targeted for the specific needs of the biologists in their laboratory. 4

5 Figure 1: A schema for seamless gene deletion Programme of work The programme of work will follow the above outlines of the methodology, a standard software engineering approach. The post doctoral researcher (Michael Riley) will spend the first three months on specification. The specification will be influenced by available open source packages, user interfaces and publications about primer design. How will look at how these, or ideas from these, can be incorporated. He will also make use of ideas from workflow systems and the construction of toolkits from other domains. The specification will be developed to describe the scope of the components of the system and the mechanisms for combination of these components. This specification will be produced as a technical report. During the next two months he will then design a library of modules. Interfaces, test suites and documentation will be planned. Decisions will be taken about the best ways to make components modular and to allow for extensibility to the toolkit. We will decide the coding conventions, and project infrastructure (version control, bug tracking, project hosting). Throughout this project Michael will be encouraged to learn and apply professional software engineering practice. This library will now be implemented. We will allow five months for implementation, according to the specification and design. Unit testing will be added during development. A further two months will be allowed for initial testing, bug fixing and reimplementation of problem areas. At the end of this stage a preliminary version will be publicly released. Extra documentation will be written at this point to encourage external users to become involved in the project. We will also now feed back our code and ideas to other projects such as Primer3. 5

6 The seamless gene deletion case study should now be achievable and we should be able to derive primers for deletions of each of S. cerevisiae s genes. This result would be published, along with a description of the toolkit. The primers will become a valuable resource for yeast biologists. The final months of this project will include time to begin to extend and support the basic library according to the wishes of external users, until it reaches a self-supporting status. Binary installation packages will be provided for a variety of platforms. The development cycle will be made stable and maintainable. The final months will also include some contingency time (as is common practice in software engineering projects). There will be weekly project management meetings, involving the postdoc and the investigators. The expertise of other members of the Robot Scientist research group will be available on request. The post doctoral researcher will be part of the wider Robot Scientist research group at Aberystwyth, a group with expertise in software, automation, algorithms, and yeast biology. Business plan A key problem with bioinformatic software is support beyond the end of the grant. This toolkit will be provided open source, and will therefore be freely available to anyone. We expect that our tool will offer substantial advantages over existing tools. Our plan to continue support and development of our tool is to use these advantages to leverage financial support from commercial users of the tool, for example by providing user support or the coding of requested interfaces for a fee. In the past we have been successful in getting such commercial support of bioinformatic programs for protein structure prediction. References [1] R. Akada et al. PCR-mediated seamless gene deletion and marker recycling in Saccharomyces cerevisiae. Yeast, 23 (5): , [2] S. Arvidsson et al. QuantPrime - a flexible tool for reliable high-throughput primer design for quantitative PCR. BMC Bioinformatics, 9:465, [3] M. Bekaert and E. C. Teeling. Uniprime: a workflow-based platform for improved universal primer design. Nucleic Acids Research, 36(10):e56, [4] J. Fredslund and M. Lange. Primique: automatic design of specific PCR primers for each sequence in a family. BMC Bioinformatics, 8:369, [5] P. M. K. Gordon and C. W. Sensen. Osprey: a comprehensive tool employing novel methods for the design of oglionucleotides for DNA sequencing and microarrays. Nucleic Acids Research, 32(17):e133, [6] W. A. Kibbe. OligoCalc: an online oligonucleotide properties calculator. Nucleic Acids Research, 35, [7] T. Koressaar and M. Remm. Enhancements and modifications of primer design program primer3. Bioinformatics, 23(10): , May [8] S. Rozen and H. J. Skaletsky. Primer3 on the WWW for general users and for biologist programmers. In Bioinformatics Methods and Protocols: Methods in Molecular Biology, pages Humana Press, Totowa, NJ, [9] F. M. You et al. BatchPrimer3: A high throughput web application for PCR and sequencing primer design. BMC Bioinformatics, 9:253,