Feature Meeting Review: 2002 O’Reilly Bioinformatics Technology Conference Westin La Paloma Resort, Tucson, Arizona, USA. January 28–31 2002. Damian Counsell* MRC UK Human Genome Mapping Project Resource Centre, Hinxton, Cambridge, CB10 1SB, UK * Correspondence to: MRC UK Human Genome Mapping Project Resource Centre, Hinxton, Cambridge, CB10 1SB, UK. E-mail: [email protected]Received: 2 April 2002 Accepted: 5 April 2002 Abstract At the end of January I travelled to the States to speak at and attend the first O’Reilly Bioinformatics Technology Conference [14]. It was a large, well-organized and diverse meeting with an interesting history. Although the meeting was not a typical academic con- ference, its style will, I am sure, become more typical of meetings in both biological and computational sciences. Speakers at the event included prominent bioinformatics researchers such as Ewan Birney, Terry Gaasterland and Lincoln Stein; authors and leaders in the open source programming community like Damian Conway and Nat Torkington; and representatives from several publishing companies including the Nature Publishing Group, Current Science Group and the President of O’Reilly himself, Tim O’Reilly. There were presentations, tutorials, debates, quizzes and even a ‘jam session’ for musical bioinformaticists. Copyright #2002 John Wiley & Sons, Ltd. Keywords: bioinformatics; open source; BLAST; visualization; programming; O’Reilly Introduction Over 700 biologists, computer scientists, bioinfor- maticists, hackers, publishers and journalists came (some at great personal expense) to Tucson, Arizona to listen, argue, share and to write computer code. In his introduction to one of the keynotes, Tim O’Reilly explained why a computer book company and documentation consultancy had organized a bio- technology conference. Last year O’Reilly published its first bioinformatics text [10]. It is not the best introductory bioinformatics text, but imprint’s reputation with the so-called ‘open source’ commu- nity was enough to make it immediately (and temporarily) Amazon’s best selling computer book. Many of O’Reilly’s loyal readers would happily refer to themselves as ‘hackers’, meaning ‘indepen- dent voluntary programmers’ rather than ‘computer criminals’. (The cognoscenti use the term ‘crackers’ to refer to the sort of people who deface Web sites or steal passwords.) Hackers lie at the heart of the enormously successful and growing open source software movement. Open source hackers write computer code and documentation to solve problems, to learn, and to impress others. They make all of their original human-readable program code available (under copyright) for improvement or modification. If others distribute programs based on the original source they are often required by the copyright- holders to make the source code of these changes available to others in turn, otherwise they are free to do what they will with it. Hackers make a distinction between this meaning of ‘free’, ‘free as in speech’, and ‘free as in beer’. This ethos and several open source licences have been adopted by most of the major bioinformatics programming projects including EMBOSS [7], Ensembl [8], BLAST [6] and the those of the Open Bioinformatics Foundation [13]: BioPerl, BioJava, BioPython and so on. Curious hackers Many non-scientists were present at this meeting. Part of the success of O’Reilly’s foray into bio- informatics publishing derives from the strong and well-intentioned curiosity of this hacker fraternity Comparative and Functional Genomics Comp Funct Genom 2002; 3: 264–269. Published online 2 May 2002 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002 / cfg.170 Copyright # 2002 John Wiley & Sons, Ltd.
Westin La Paloma Resort, Tucson, Arizona, USA. January 28–31 2002.
Damian Counsell*MRC UK Human Genome Mapping Project Resource Centre, Hinxton, Cambridge, CB10 1SB, UK
*Correspondence to:MRC UK Human GenomeMapping Project Resource Centre,Hinxton, Cambridge, CB10 1SB,UK.E-mail:[email protected]
Received: 2 April 2002
Accepted: 5 April 2002
Abstract
At the end of January I travelled to the States to speak at and attend the first O’Reilly
Bioinformatics Technology Conference [14]. It was a large, well-organized and diverse
meeting with an interesting history. Although the meeting was not a typical academic con-
ference, its style will, I am sure, become more typical of meetings in both biological and
computational sciences.
Speakers at the event included prominent bioinformatics researchers such as Ewan Birney,
Terry Gaasterland and Lincoln Stein; authors and leaders in the open source programming
community like Damian Conway and Nat Torkington; and representatives from several
publishing companies including the Nature Publishing Group, Current Science Group and
the President of O’Reilly himself, Tim O’Reilly. There were presentations, tutorials,
debates, quizzes and even a ‘jam session’ for musical bioinformaticists. Copyright #2002
John Wiley & Sons, Ltd.
Keywords: bioinformatics; open source; BLAST; visualization; programming; O’Reilly
Introduction
Over 700 biologists, computer scientists, bioinfor-maticists, hackers, publishers and journalists came(some at great personal expense) to Tucson, Arizona tolisten, argue, share and to write computer code. In hisintroduction to one of the keynotes, Tim O’Reillyexplained why a computer book company anddocumentation consultancy had organized a bio-technology conference. Last year O’Reilly publishedits first bioinformatics text [10]. It is not the bestintroductory bioinformatics text, but imprint’sreputation with the so-called ‘open source’ commu-nity was enough to make it immediately (andtemporarily) Amazon’s best selling computer book.
Many of O’Reilly’s loyal readers would happilyrefer to themselves as ‘hackers’, meaning ‘indepen-dent voluntary programmers’ rather than ‘computercriminals’. (The cognoscenti use the term ‘crackers’to refer to the sort of people who deface Web sitesor steal passwords.) Hackers lie at the heart of theenormously successful and growing open sourcesoftware movement.
Open source hackers write computer code and
documentation to solve problems, to learn, and toimpress others. They make all of their originalhuman-readable program code available (undercopyright) for improvement or modification. Ifothers distribute programs based on the originalsource they are often required by the copyright-holders to make the source code of these changesavailable to others in turn, otherwise they are free todo what they will with it. Hackers make a distinctionbetween this meaning of ‘free’, ‘free as in speech’, and‘free as in beer’.
This ethos and several open source licences havebeen adopted by most of the major bioinformaticsprogramming projects including EMBOSS [7],Ensembl [8], BLAST [6] and the those of the OpenBioinformatics Foundation [13]: BioPerl, BioJava,BioPython and so on.
Curious hackers
Many non-scientists were present at this meeting.Part of the success of O’Reilly’s foray into bio-informatics publishing derives from the strong andwell-intentioned curiosity of this hacker fraternity
Comparative and Functional Genomics
Comp Funct Genom 2002; 3: 264–269.Published online 2 May 2002 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002 /cfg.170
Copyright # 2002 John Wiley & Sons, Ltd.
about the maturing field of bioinformatics. Hackersare particularly annoyed, if not always so wellinformed, by any restriction being placed on theavailability of sequence data. They see genomicdata as software and are disturbed by any attemptto ‘close its source’. They see parallels betweensoftware patents (common in the States, but notpossible in Europe) and ‘gene patents’.
That actual computer programming was a part ofthe proceedings is testimony to the meeting’sunusual background. The ‘Bio Hackathon’ at theconference was a successful effort in ‘live’ softwaredevelopment by various programming groups fromthe Open Bioinformatics Foundation. It began theweekend preceding the week of the meeting andcontinued in Cape Town, South Africa. I mustconfess that I turned down the chance to participatein this (in place of Alan Bleasby on behalf of theEMBOSS team). Sadly I’m not the sort of pro-grammer who can write substantial programs indays, but I was happy to accept an invitation tospeak on the Bioinformatics.Org [3] track.
Bioinformatics.Org
Bioinformatics.Org is a non-profit academic organ-ization, established in 1998 at the University ofMassachusetts Lowell. As well as providing a vir-tual home for a large number of open sourcebioinformatics projects it campaigns for freedomof information in the biosciences. It is also inter-ested in bioinformatics education. It was true to thespirit of the event that Bioinformatics.Org tookover a lecture theatre for its own one-day track.
Bioinformatics.Org’s Executive Director, JeffBizarro, took the opportunity of the O’Reillymeeting – the Second Annual Meeting of Bioinfor-matics.Org – to present the organization’s 2002Benjamin Franklin Award to Michael B. Eisen ofthe Lawrence Berkeley National Laboratory and theUniversity of California at Berkeley. This award wasdoubly appropriate, as Eisen wrote ScanAlyze,Cluster and TreeView, immensely popular softwarefor cluster analysis of microarray data, and is alsoone of the principal moving forces behind thePublic Library of Science [16] movement.
The organization of the conference
It would be impossible to cover all of the talks indetail. I followed my own personal path through the
programme and here I report on some of the pre-sentations that made the biggest impression on me.
Presumably to cater for such a broad range ofattendees, the conference was sensibly organizedalong three concurrent technical/academic tracksand one commercial track. I did, however, hearcomplaints (compliments?) from delegates that toomany interesting events coincided.
The ‘Fundamentals’ track was aimed at bothbiological and computational beginners. Though Icannot comment on the biological talks on the‘Analysis’ track, it certainly carried most of themore heavily computational talks I heard. Mostpractically oriented was the ‘Discovery’ track whichalso carried a couple of presentations about ethicalissues in bioinformatics, issues discussed withvigour both inside and outside sessions.
The presentations were further subdivided by styleinto keynotes, tutorials (with accompanying boundtexts) and other, more informal, gatherings (birds-of-a-feather sessions and community meetings).
‘BLAST programming’ – Thomas Madden
Even the least computer-literate biologist will,knowingly or not, have used a BLAST search toolto find identical or similar sequences in the ever-growing genomic databases. Two of the best pre-sentations I attended centred on this marvelloustool. Thomas Madden is one of the core BLASTdevelopers at the NCBI. For me his talk justifiedmy journey on its own. I am currently planning acollaborative bioinformatics system in which astandalone BLAST server will play a central part.
This tutorial made clear – albeit in fairly tech-nical terms – that BLAST is more than just themost commonly used bioinformatics program. Ithas developed into a comprehensive and customi-sable platform with its own storage and outputformats, programming interfaces and architecture.If the audience was anything to go by, there is alsoan active community of programmers and adminis-trators who not only use BLAST, but also havecreated new variants of the original code.
‘Linux clusters for bioinformatics’ – GlenOtero
Unlike other, proprietary computer operating sys-tems, running Linux on one hundred PCs costs no
Meeting Review 265
Copyright # 2002 John Wiley & Sons, Ltd. Comp Funct Genom 2002; 3: 264–269.
more than running Linux on one PC. Even moresurprising, one copy of Linux might well cost younothing, if you download a so-called ‘distribution’ –a pre-packaged collection of kernel (the coreoperating system), installation tools, documentationand software from the Net. It’s usually easier,however, to buy a boxed Linux distribution withCDs and manuals. Even this is cheaper than atypical shrinkwrapped edition of Windows.
Linux was not only built across the Net byremotely collaborating programmers, it is also anetwork-centric operating system. These two attri-butes are probably connected. Because Linuxcombines sensible licensing with fast, refined andintegrated network capabilities it is perfect for com-bining computers locally in parallel via Ethernet.Such ‘clusters’ of cheap, generic machines can pooltheir resources to solve computational problemsthat could not be tackled by standalones. Thepower of desktop processors of the Pentium andAthlon lines improves so rapidly and the machinesbased on these chips are such commodity items thatthe price/performance characteristics of these sys-tems can overtake those of slower-developing super-computer architectures. Linux clusters of these andother processors have become popular with manyacademics both as tools and as objects of study inthemselves. Glen Otero of Linux Prophet [15] des-cribed the application of this technology to bioin-formatics.
This was the second tutorial I attended on thefirst day and any learning-fatigue I might have beensuffering had dispersed at the start of Otero’senthusiastic and jokey presentation. Offering freesoftware CDs for anyone who could use one of acollection of improbable words in a question,juggling with illuminated balls and mocking bothhimself and his audience Dr. Otero (‘don’t be putoff by the ‘Doctor’’) gave an extraordinarilycomprehensive introduction to the practicalities ofchoosing and running a Linux cluster in a biologicalenvironment. Otero is a consultant to biologicalresearchers thinking of setting up such ‘discountsupercomputers’.
‘Open source bioinformatics’ – EwanBirney
Ewan Birney is the young and charismatic leader ofthe Ensembl group at the European BioinformaticsInstitute. He has always been an enthusiastic
advocate of Open Source, both sharing genomicdata and sharing bioinformatics source code. Hisfunny and relaxed presentation chimed well withthe feelings of the attendees. He joked about themisunderstandings between biologists and compu-tational scientists (over the meanings of words like‘vector’, for example). He talked about his ownwork on the Ensembl project and with the variousOpen Bioinformatics Foundation Open Sourceprojects: BioPerl, BioJava, BioPython, BioCorba,BioDAS.
He also made ‘the case for bioinformatics’. Whilebioinformaticians are often frustrated by the reluc-tance of the biological establishment to embracecomputational tools for biology, his belief that‘bioinformatics still hasn’t been hyped enough’caused me a certain amount of discomfort. Tenyears ago, however, people were equally skepticalabout the Human Genome project and the prospectsfor the cloning of large mammals; I hope to beproved a pessimist about the potential of the field.
‘Interactive data visualization’ – JohnHotchkiss
The original speaker planned for this presentationwas the founder of AnVil Informatics, Inc. (AVI),Georges Grinstein of the University of Massachusettsat Lowell. Unfortunately he lost his voice before hecould talk, but John Hotchkiss, Chief TechnologyOfficer of AnVil Informatics, bravely stepped for-ward to give us a brisk tour of a whole range oftechnologies – doubly brave since he was presentingsomeone else’s slides. He did an excellent job.
Genomics produces vast quantities of data.Humans are notoriously poor at absorbing andretaining individual items of information, butfamously good at identifying patterns. Hotchkissbegan with John Snow’s simple street map plottingof cholera deaths around a water pump in London.The Broad Street pump cholera outbreak of 1854is now an exemplar in epidemiology. Hotchkisspointed out that, contrary to the popular image,this map was more a tool of argument than one ofinvestigation. It still makes a convincing case today.This important distinction led to the argument thatvisualization approaches could usefully be dividedaccording to the kind of interaction users madewith the data being processed. Some applications ofvisualization were for exploratory, some for con-firmatory and some for production purposes.
266 Meeting Review
Copyright # 2002 John Wiley & Sons, Ltd. Comp Funct Genom 2002; 3: 264–269.
As he proceeded, Hotchkiss introduced some ofthe interesting techniques developed inside andoutside his company to handle and clarify multi-dimensional data of the kind commonly encoun-tered in genomics. Some of these techniques werestrikingly simple and effective. He showed, forexample, that substituting complex icons forcoloured dots could separate different populationsof data points in plots even more clearly. One ofAVI’s own proprietary approaches, RadViz, simplymaps values onto radii of a circle, but, morecleverly, maximizes the usefulness of this approachby optimising the arrangement of those spokes forrevealing patterns.
‘Data visualization for genomics’ –Timothy M. Kunau
Timothy Kunau of the University of Minnesotapursued some of these themes further, although hisfocus initially was more on tools for developmentsuch as ‘Integrated SYStem’ (ISYS), a set of com-ponents developed by the National Center forGenome Resources (NCGR) for the exploration ofgenomic data. ISYS is based on the establishedJava/Swing programming toolkit. This offered, forexample, a metabolic pathway viewer. ISYS seemedto be a perfect example of a tool designed withsharing in mind. Different developers in differentlabs could produce components independently, butthose components had a similar look and feel.
Kunau outlined the ‘programming by cartoon’approach of University of Pennsylvania’s bio-Widgets [5] toolkit. Users can build their ownbioinformatics pipeline by assembling visual repre-sentations of various modules. For example multi-ple simultaneous comparisons could be performedby connecting a single sequence input to variousprocessing modules.
He also discussed the visualization techniquesused in MetaFam [12], an interfamily protein brow-ser to represent not just the existence of connectionsbetween various proteins, but the strengths andnature of those relations.
Finally Kunau addressed the simple practical issuesof screen dimensions. You can see more if your screenis bigger. You can see more if you can see in threedimensions. He showed (two-dimensional) slides ofa system called ‘GeoWall’ in action. Based on tech-nology originating at the Electronic VisualizationLaboratory of the University of Illinois at Chicago,
the GeoWall [9] consortium, including membersfrom the Universities of Minnesota and Michigan,have developed a means of using PCs or Macs toput three dimensions of data onto lecture theatreviewing screens. The system was built with visuali-zation of geological data in mind, but applicationsto bioinformatics were discussed.
‘Speedup at what cost?: Heuristic vs.complete algorithms in homologysearch’ – Christopher Dwan
This was one of those rare computational talksaimed at the biological contingent and an out-standing example of the art of explanation. Itinspired me to revise my own approach to teachingbioinformatics. Christopher Dwan of University ofMinnesota’s Center for Computational Genomics andBioinformatics gave the clearest explanation ofdistinctions and choices to be made in practicalbioinformatics (and applied computing in general)that I have ever seen.
He did two difficult things well: explained thespecific functional differences between two of themost important sequence alignment algorithms andtheir consequences for practising researchers, andexplained the general distinctions computer scien-tists make between different classes of problem-solving algorithm.
In this case the two algorithms compared werethe complete Smith-Waterman and heuristic BLASTsequence search (alignment) methods. The simpleconclusion of this talk was that BLAST is quicker/cheaper and misses some matches while SmithWaterman is slower/more expensive and finds allmatches (at least all those possible given a specificscoring system and cut-off). Not only is this a grossover-simplification of Dwan’s presentation, but itdoes no justice to the elegant way Dwan combinedempirical demonstration – using data obtained dur-ing the conference itself with an exhibitor’s system –with good, old-fashioned exposition.
‘Computing Strategies for theinterpretation of mass spectral data forproteomics applications’ – WilliamGleason
William Gleason [11] (University of Minnesota) gavea witty and insightful talk about some of the most
Meeting Review 267
Copyright # 2002 John Wiley & Sons, Ltd. Comp Funct Genom 2002; 3: 264–269.
advanced proteomics methods. As biologists drownin nucleotide sequences he emphasised the import-ance of proteomics to biology and quoted GregPetsko in describing proteomics as ‘going fromsequence to consequence’.
With the advent of what he referred to as ‘soft’fragmentation techniques such as MALDI (MatrixAssisted Laser Desorption Ionization) and ESI(ElectroSpray Ionization) it has become possible tobreak proteins up into analysable ions with muchless damage. If a technique such as ESI is coupledto capillary electrophoresis or high-pressure liquidchromatography (HPLC) then sequencing of thesefragments can be done very rapidly on tiny(femtomolar) quantities of material.
Gleason described how a cluster of Linux PCs(the University of Minnesota SupercomputingInstitute Netfinity Linux Cluster) running theLutefisk1900 and CIDentify software packagescould be used to rapidly analyse the fragmentarysequence output from experiments such as these.His goal was to obtain answers sufficiently quicklyfor the settings of the mass spectrometer to beadjusted in order to maximise the usefulness of itsoutput.
‘Project management atBioinformatics.Org’ – Gary vanDomselaar
As well as being on the Executive Committee ofBioinformatics.Org, Gary van Domselaar adminis-ters Bioinformatics.Org’s computer systems. He is abioinformatics scientist at the Genetics Institute anda PhD. Candidate in David Wishart’s researchgroup in the Faculty of Pharmacy and Pharmaceu-tical Sciences in Edmonton. He gave a revealing talk,both about the logistics of hosting a wide range ofbiological computing projects distributed aroundthe globe, and about the nature of some of theprojects. Even as a regular visitor and contributorto the Bioinformatics.Org site I learned aboutseveral features which were new to me.
‘DHTML and scalar vector graphics inbioinformatics’ – Malay Kumar Basu
One of these revelations was a gem of a projectcreated by Malay Kumar Basu in ‘downtime’ fromhis full-time study. Basu is a graduate student in
Molecular Biology in the Centre for Cellular andMolecular Biology, India. SeWeR (Sequence analy-sis Web Resources) [18] is his ingenious and simpleWeb interface to an array of server-based bioinfor-matics programs. It allows a naive user to applydatabase search, sequence analysis and even visua-lization programs to his or her data in a simple andcustomisable way.
Basu’s striking talk described the philosophy andtechnology behind the system. SeWeR is a perfectexample of some of the extraordinary individualefforts taking place in the open source softwaredevelopment community. He also spoke aboutsome of his more recent projects, including a Perllibrary for the generation of scalable vector gra-phics (SVG) and derived modules for visualizingbiological data.
‘Using the NCBI C++ toolkit in thedevelopment of the BIND database’ –Doron Betel
This talk was oriented more strongly towards pro-grammers than any other I attended, with detailedsource code examples displayed throughout. TheNCBI toolkit is an extraordinarily wide-rangingcollection of C++ code for bioinformatics soft-ware development. In its functionality it rivals thelibraries underlying the European Molecular Bio-logy Open Software Suite (EMBOSS). It is a testi-mony to its accessibility that Doron Betel producedhis BIND database for the manipulation of bio-chemical pathway data completely independently ofthe toolkit’s authors. He is a graduate student in theChris Hogue’s Bioinformatics Lab at the Mt. Sinai
Hospital Research Institute in Toronto; as he put it:‘I’m not at the NCBI and I’ve never even beenthere’.
While I was impressed by the vast range offunctions available in the kit, its code-generationutility (programs which do their own programmingwill always be popular with programmers), its docu-mentation and its cross-platform nature, I began tobe a little daunted by the level of abstraction atwhich it operates. For example an elaborate HTMLpage with multiple hierarchical components can begenerated in the toolkit by a single ‘print’ com-mand, but a firm grasp of such structures is neces-sary to use this kind of power sensibly.
268 Meeting Review
Copyright # 2002 John Wiley & Sons, Ltd. Comp Funct Genom 2002; 3: 264–269.
‘BioMail: as an example of pushtechnology in bioinformatics’ – DmitryMozzherin
BioMail [4] is a classic example of a bioinformaticsresource which has become a raging success becauseit is useful, simple and free. Dmitry Mozzherin des-cribed how his easy-to-use email publication alertsystem had caught on with individual biologists andsuperseded expensive commercial services in severaluniversity libraries because of its convenience andreliability.
The system is hosted at Bioinformatics.Org.
Conclusion
The O’Reilly Bioinformatics Conference managedto be both scholarly and lively. I learned a lot; Idiscovered bioinformatics projects I had not pre-viously encountered and understood more aboutfamiliar projects. While the meeting dealt withscience and technology, it also had a ‘philosophy’.
Computer programming was originally consid-ered something of an academic pursuit. As itbecame a big business, the previous science-likeethic of sharing code disappeared. The rise of theNet has revitalized this ‘collegiate’ approach andinspired the open source movement and its frater-nity of self-proclaimed hackers. It would be easy todismiss them as a semi-anarchic rabble if not for thedazzling successes of the products of their colla-borations – including the infrastructure of the Netitself. Programs such as Linux, the free operatingsystem, Apache [1], the most popular Web servingsoftware, BIND [2], the code which assigns iden-tities to almost every internet-connected device, andsendmail [17], which handles the vast majority ofemail, are all open source creations. In bioinfor-matics, of course, a great deal of open source codewas used to map, sequence and assemble the humangenome.
Now that biology and computing are convergingit is naturally members of this open source commu-nity and who are most eager to bring the philoso-phy of shared enterprise back to the scientific worldwhence some feel it came. O’Reilly, as court pub-lishers to the hacker nation may have become acci-dental pioneers of a new kind of scientific gathering.It is likely that future biotechnological meetings willalso be more open to intelligent ‘outsiders’, moreconcerned with explanation and more fun.
The Meeting Reviews of Comparative and Func-tional Genomics aim to present a commentary onthe topical issues in genomics studies presented at aconference. The Meeting Reviews are invited; theyrepresent personal critical analyses of the currentreports and aim at providing implications for futuregenomics studies.
