Proc. Natl. Acad. Sci. USAVol. 90, pp. 4338-4344, May 1993

Review

The human genome projectMaynard V. OlsonDepartment of Molecular Biotechnology, University of Washington, Seattle, WA 98195

ABSTRACT The Human GenomeProject in the United States is now wellunderway. Its programmatic direction waslargely set by a National Research Councilreport issued in 1988. The broad frame-work supplied by this report has survivedalmost unchanged despite an upheaval inthe technology of genome analysis. Thisupheaval has primarily affected physicaland genetic mapping, the two dominantactivities in the present phase ofthe project.Advances in mapping techniques have al-lowed good progress toward the specificgoals of the project and are also providingstrong corollary benefits throughout bio-medical research. Actual DNA sequencingof the genomes of the human and modelorganisms is still at an early stage. Therehas been little progress in the intrinsicefficiency of DNA-sequence determination.However, refinements in experimental pro-tocols, instrumentation, and project man-agement have made it practical to acquiresequence data on an enlarged scale. It isalso increasingly apparent that DNA-sequence data provide a potent means ofrelating knowledge gained from the studyof model organisms to human biology.There is as yet little indication that theinfusion of technology from outside biologyinto the Human Genome Project has beeneffectively stimulated. Opportunities in thisarea remain large, posing substantial tech-nical and policy challenges.

In the United States, the Human GenomeProject first took clear form in February1988, with the release of the NationalResearch Council (NRC) report Mappingand Sequencing the Human Genome (1).To a degree remarkable in Federal sci-ence policy, this report has had a cleareffect on subsequent programmatic ac-tivity. With a budget in the current fiscalyear of $170 million, jointly administeredby the- National Institutes of Health(NIH) and the Department of Energy(DOE), a program is underway that con-forms closely to the recommendations ofthe NRC committee. After a 5-year real-ity check, it is of both scientific andpolicy interest to examine how the com-mittee's view toward this project hasfared in the field.

Background

The human genome is the genetic mate-rial in human egg and sperm cells (i.e.,

germ cells), which contain 3 x 109 basepairs (bp) of DNA. Given the four-letteralphabet of DNA-customarily symbol-ized with the letters G, A, T, and C-thesequence of 3 x 109 bp corresponds to750 megabytes of information. If the se-quence of the human genome could bedetermined, it would be possible to storeand manipulate it on a desktop computer.However, even the dream of acquiringDNA sequence on this scale is of recentorigin. Dramatic progress was made dur-ing the 1950s and 1960s in understandingthe mechanisms by which genetic infor-mation specifies biological structure andfunction. However, during this era, theinformation itself remained nearly inac-cessible.A landmark event in DNA analysis

came in 1970 with the discovery of site-specific restriction enzymes (refs. 2 and3; ref. 4, p. 64). These remarkable en-zymes have the ability to scan any sourceof DNA for every occurrence of a par-ticular string of bases-for example, theenzyme EcoRI recognizes the stringGAATTC. Restriction enzymes cleaveboth strands of the double helix at theirrecognition sites. Since the cleavageevents are directed by the DNA se-quence, they always occur at the samepositions in different samples of genomicDNA extracted from any genetically ho-mogeneous source (e.g., different tissuesof the same individual human or differentindividuals sampled from an inbred strainof mice).

Restriction enzymes provide a meansto develop precise physical maps ofDNAsimply by determining the coordinates inbase pairs of the sites at which particularenzymes cleave (ref. 4, pp. 66-67; ref. 5).Like topographic maps, physical maps ofDNA derive their utility through annota-tion: mapped landmarks provide refer-ence points relative to which functionalDNA sequences such as genes can belocalized. Restriction enzymes also facil-itate a key step in the cut-and-spliceprocedures by which recombinant-DNAmolecules (i.e., DNA clones) are con-structed (ref. 4, pp. 73-74 and pp. 99-124).The importance of recombinant-DNA

technology is often attributed primarilyto its synthetic dimension. For example,the ability to design and construct aDNAmolecule that programs a bacterium to

synthesize a mammalian protein providesa route to large amounts of the pureprotein. The ability to alter the structureof the protein through site-directed mu-tagenesis lends genuine novelty to theresultant biosynthetic opportunities.However, the importance of recombi-nant-DNA technology in making the Hu-man Genome Project feasible stems fromits analytical dimensions. Cloning pro-vides a means to purify individual recom-binant-DNA molecules from complexmixtures and then to prepare biochemi-cally useful amounts of the molecules byculturing the microbial strains into whichthey have been introduced.A less obvious consequence of the

discovery of restriction enzymes was thedevelopment of the first practical methodof genetic mapping in humans (ref. 4, pp.519-522; ref. 6). Most human cells con-tain two copies of each DNA sequence,one of maternal and the other of paternalorigin. When a new germ cell is pro-duced, it contains only one copy of thegenome, a copy that is a unique mosaic ofthe two genomes from which it was de-rived. Genetic mapping involves measur-ing, through actual inheritance studies infamilies, the probability that two closelyspaced segments of the genome will staytogether during germ-cell formation. Themapping requires an ability to distinguishbetween the two copies of the genomepresent in the somatic cells from whichthe germ cells are derived. Subtle differ-ences in the base sequence of differentinstances of the human genome some-times alter restriction sites and, hence,restriction-fragment sizes. These alter-ations are detectable even in complexgenomes by a method known as gel-transfer hybridization, which was devel-oped in 1975 (ref. 4, pp. 127-130; ref. 7).In 1987, the first global human geneticmap, based on "restriction-fragment-length polymorphisms," was published(8).As to the actual determination ofDNA

sequence, reasonably efficient methodsfirst appeared in 1977 (ref. 4, pp. 67-69;refs. 9 and 10). A technique known as

Abbreviations: DOE, Department of Energy;FISH, fluorescence in situ hybridization;NIH, National Institutes of Health; NRC,National Research Council; STS, sequence-tagged site; YAC, yeast artificial chromo-some.

4338

Proc. Natl. Acad. Sci. USA 90 (1993) 4339

chain-termination sequencing came todominate standard practice: it is based onenzymatic DNA synthesis, carried out invitro in the presence of artificial chain-terminating variants of the normal DNA-precursor molecules. By the early 1980sindividual sequences exceeding 105 bphad been determined (11); however, amore common scale of analysis was 103-104 bp. Most physical mapping was car-ried out on a similar scale.Given the gap between the ability to

determine 105 bp of DNA sequence in astate-of-the-art laboratory and the 109-bpsize of the human genome, the NRCcommittee confronted an enormousproblem of scale. Partly because of theobvious need for improved technologyand also because of a desire to maximizesynergy between genome analysis andstudies of biological function, the com-mittee recommended against early em-phasis on large-scale sequencing of hu-man DNA. Instead, it advocated com-prehensive physical and genetic mappingof the human genome, extensive mappingand sequencing ofthe smaller genomes ofseveral model organisms, and a system-atic effort to develop improved sequenc-ing technology.

Principal Aims of the HumanGenome Project

More important than the specific map-ping and sequencing objectives of theHuman Genome Project are threebroader aims that are implicit in thesegoals:

(i) To improve the research infrastruc-ture of human genetics.

(ii) To help establish DNA sequence asthe primary interface between knowledgeof human biology and knowledge of thebiology of model organisms.

(iii) To launch an open-ended effort toimprove the analytical biochemistry ofDNA.For the purposes of this review, prog-

ress in the Human Genome Project willbe examined relative to these three broadaims.

The Research Infrastructure ofHuman Genetics

In the context of the Human GenomeProject, research infrastructure refers tothe biological, informational, and meth-odological tools with which genetics re-search is carried out. Intensive geneticanalysis of any species is heavily depen-dent on infrastructure. Particularly im-portant are genetic-linkage maps, physi-cal maps of DNA, and characterizedDNA clones. The latter are useful as

reagents that can be used to assay forparticular short segments of the genomeby DNA-DNA hybridization (ref. 12, pp.

188-191) and as the starting points forsequence analysis or functional studies.Human genetics is uniquely dependent

on strong research infrastructure. Whilemodel organisms have been extensivelybred for the specific purpose of facilitat-ing genetic analysis (13), human geneticsis limited to the examination of individ-uals, families, and populations as theyare found in contemporary society.Hence, the NRC committee set ambi-tious goals for the construction of de-tailed physical and genetic maps of thehuman genome, as well as organized col-lections of cloned human DNA. By de-sign, the goals were too ambitious for thetechnology of 1988. In retrospect, theywere so ambitious that they probablywould have overwhelmed the basic meth-odologies on which the NRC report wasbased. Fortunately, technical advancessince 1988 have exceeded all reasonableexpectations.Much of this progress was made pos-

sible by the development of the polymer-ase chain reaction (PCR). PCR, which isessentially a method of in vitro cloning,allows the amplification of specific DNAmolecules in vitro through cycles of en-zymatic DNA synthesis (ref. 4, pp. 79-85). PCR amplification is dependent on apair of short, synthetic "primers" (i.e.,single-stranded DNA molecules whoseends can be extended by DNA polymer-ase under the direction of template mol-ecules). The test sample provides thetemplate molecules, and the primers di-rect the amplification to a particular seg-ment of the template DNA, typically aregion only a few hundred base pairs inlength. Starting with a minute sample oftotal human DNA, it is possible to am-plify any such region 1 billionfold whileleaving the rest of the genome at itsoriginal concentration.Widespread application of the PCR de-

pends on an efficient, automated methodfor the chemical synthesis of the PCRprimers. An approach to DNA synthesisbased on phosphoramidite chemistry,which became routine in the early 1980s,meets this need (ref. 4, pp. 69-70; refs.14-16). The first paper on the PCR ap-peared in 1985 (17) but received littlenotice; for example, despite its presentprominence in genome analysis, it is notmentioned in the NRC report. The ex-plosive growth of PCR applications be-gan with the publication of an importantrefinement ofthe PCR protocol in 1989-the use of a thermostable DNA polymer-ase (18). This refinement allowed thecycles ofDNA synthesis, which are anal-ogous to cellular generations, to bedriven by simple thermal cycling with no

new addition of reagents at each cycle.By the end of 1989, it was already

apparent that the PCR provided a prac-tical means of abstracting large-scalephysical maps away from the particular

methods used to construct them. Thiscapability required a new choice of land-marks called sequence-tagged sites(STSs; ref. 19). An STS is simply a short,unique sequence of DNA that can beamplified via the PCR. STSs are ideallandmarks during map construction be-cause of the ease with which they can bedetected by PCR assays. Equally impor-tant is their role in map representationand map use.Complex physical maps based on re-

striction sites are of little value as exper-imental tools unless they are supportedby a collection of clones that can be usedto detect particular segments of themapped DNA via DNADNA hybridiza-tion assays. Comprehensive maps of thehuman genome would have to be sup-ported by tens of thousands of clones,each of which would have to be main-tained as a separate microbial strain. Incontrast, STSs can be described in anelectronic data base in a form that makesthem experimentally accessible in anylaboratory. The most critical aspect ofanSTS description is the DNA sequence ofthe two primers. Laboratory implemen-tation of an STS simply requires that thetwo primers be synthesized and the ap-propriate temperature-cycling regime becarried out.Most large-scale physical maps are

constructed through the process of "con-tig building." A contig is an organized setof DNA clones that collectively provideredundant cloned coverage of a regionthat is too long to clone in one piece (ref.4, pp. 587-588; refs. 20-22). Typically,the clones have random end points, andthe contig is described by specifying theamount of overlap between each cloneand its nearest neighbors. A procedurereferred to as STS-content mapping pro-vides a convenient method of establish-ing these overlaps (ref. 4, pp. 610-612;ref. 23). In a step that precedes contigbuilding, the STSs are tested to confirmthat they occur in a single copy in thegenome; then, if two clones share even asingle STS, they can be reliably assumedto overlap.Although the PCR has had a profound

effect on physical mapping, other newdevelopments have also improved theprospects for the construction of large-scale physical maps. One such develop-ment has been the introduction of theyeast artificial-chromosome (YAC) clon-ing system, fist described in 1987 (ref. 4,pp. 590-592; ref. 24). YACs allow largesegments ofDNA to be cloned as linear,artificial chromosomes into the yeasthost Saccharomyces cerevisiae. Evensome of the earliest YAC clones were 10times the size of the largest clones thathad been constructed previously. Fur-thermore, the YAC system appears ca-pable of cloning a higher proportion ofthe genomic DNA of many organisms

Review: Olson

Proc. Natl. Acad. Sci. USA 90 (1993)

than could be recovered using earliersystems. This point has been most clearlydocumented during the physical mappingof the genome of the nematode wormCaenorhabditis elegans (25).By 1989, YAC technology had evolved

to the point where specific segments ofthe human genome could be recoveredefficiently in YAC clones (26). Soonthereafter, multi-megabase-pair contigsbegan to appear (23, 27-29), and, in thefall of 1992, complete YAC-based phys-ical maps of human chromosome 21 (30)and the human Y chromosome (31) werepublished. In these projects, contig con-struction was largely by STS-contentmapping. There is little doubt that thesame technology employed on chromo-somes Y and 21, as well as on a largesegment of the X chromosome (29), hassufficient power to produce highly con-nected physical maps ofthe entire humangenome.Another important advance in physical

mapping has been the development offluorescence in situ hybridization (FISH)into a routine procedure. This techniqueemploys DNA probes that can detectsegments of the human genome byDNADNA hybridization on samples oflysed metaphase cells prepared underconditions that preserve the morphologyof the condensed human chromosomes.Attachment of fluorescent molecules tothe probe DNA allows visualization inthe light microscope of the position on achromosome to which the probe binds.The technique is a refinement ofpreviousin situ hybridization methods that de-pended on radiolabeling of the probesand autoradiographic detection (32). Theincreases in convenience, reliability, andresolution that have accompanied non-isotopic detection have transformed therole of in situ hybridization in physicalmapping. The first nonisotopic visualiza-tion of single-copy sequences in humanchromosomes by in situ hybridizationwas published in 1985 (33). Fluorescencedetection of single-copy sequences wasintroduced in 1987 (34), after which ap-plications expanded rapidly (35-37).FISH contributes to two aspects of

long-range physical mapping. First, it al-lows individual clones to be mapped at acoarse level long before contig building iscomplete, thereby providing reagents ofimmediate use in the analysis of targetedregions. Second, contig maps have dis-continuities whenever a site in the ge-nome is missing from the available clonecollections. FISH provides a way to or-der and orient contigs along a chromo-some even when occasional discontinui-ties exist. Early efforts to construct phys-ical maps ofhuman chromosomes, whichdepended on cosmid clones that are prop-agated in Escherichia coli, yielded rela-tively small contigs separated by discon-tinuities (38, 39). YAC-based methods

have led to greatly improved continuity,but the need remains for supplementarymethods to define the order and orienta-tion of disconnected contigs along chro-mosomes.

Radiation-hybrid mapping, which in-volves fragmentation of chromosomes incultured cells with high doses of x-raysfollowed by incorporation of the frag-ments into stable cell lines, provides stillanother solution to this problem (ref. 4,pp. 608-609; ref. 40). Current protocolsfor radiation-hybrid mapping are notablefor their abandonment of the traditionalgoal of isolating a single short segment ofthe human genome in each rodent cellline. Nearly all of the radiation-hybridlines produced by these protocols con-tain many unrelated segments of the hu-man genome. Proximity of two STSs, orother markers, is inferred by statisticalanalysis of the pattern in which theyoccur in a large collection of cell lines.Closely spaced markers have a higherprobability of occurring together in thesame cell line than do pairs of markersthat are on different chromosomes or arefar apart on the same chromosome.

While the PCR, together with such newtechniques as YAC cloning, FISH, andradiation-hybrid mapping, has led to asurge of success in physical mapping,PCR-based methods have also trans-formed genetic mapping. In particular,the PCR has allowed development of anew class of genetic markers that have aparticularly high probability of existing inalternate forms in different instances ofthe human genome.These markers are based on short,

repetitive DNA sequences that arewidely distributed in the human genome.A particularly common motif is ...

(CA), .... At sites where this motifoccurs, n, the number of repetitions ofthe dinucleotide CA, is highly variablefrom one instance of the human genometo the next (41, 42). Different values of nlead to PCR-amplification products ofdifferent lengths when the entire ...

(CA)n ... tract is amplified by usingprimers that flank the repeat; these dif-ferences are readily detected by gel elec-trophoresis. An attractive feature ofPCR-detectable genetic markers is thatthey are simply a special type of STS. Assuch, they can be readily included aslandmarks in physical maps, as well asgenetic maps, thereby providing a simplemethod ofinterrelating these two types ofmaps. Many PCR-detectable geneticmarkers have been integrated into preex-isting maps ofthe human, greatly improv-ing these maps (43). Still more recently, ahuman genetic map that is completelybased on PCR-detectable markers hasbeen constructed (44). Markers of thesame type have also transformed geneticmapping in the mouse, whose genome isthe same as that of the human (45).

A key test of the effectiveness of theinfrastructure-building features of theHuman Genome Project is the extent towhich its components are being usedeven before genome-wide physical mapsare available. The most critical test in-volves projects directed at the "position-al cloning" of genes associated with her-itable diseases. Positional cloning is astrategy that was developed during the1980s to allow determination of the bio-chemical basis of the many heritable dis-eases whose analysis has resisted themore traditional approach of direct bio-chemical analysis of diseased tissue (46).In general, the biochemical analysis ofdiseased tissue is rarely effective unlessthe genetic defect alters a protein whosemetabolic role in normal tissue is alreadyunderstood. Few of the heritable dis-eases that cause mental retardation, psy-chosis, congenital malformation, malig-nant tumors, and other similarly complexeffects meet this criterion.The first step in positional cloning is to

localize the "disease" gene by carryingout genetic mapping studies on familieswith multiple affected members. Studiesof the coinheritance of the disease withgenetically mapped DNA markers allowdetermination of the position of the genein the genome. Actual biochemical iden-tification of the gene still remains a for-midable task since the resolution of ge-netic maps in the human is rarely betterthan 1 megabase pair (Mbp). Physicalmapping and functional studies on thecloned DNA are required to find the genewithin the candidate region.

Better physical mapping methods, par-ticularly the combination ofYAC cloningand FISH analysis, have improved theprospects for positional cloning. An ex-emplary baseline case, publishedjust be-fore either of these techniques becamewidely available, is cystic fibrosis. Finalsuccess in the positional cloning of thecystic fibrosis gene required heroic phys-ical mapping efforts that never achievedany semblance ofcontinuous cloned cov-erage of the candidate region (47). Piece-meal cloning and mapping proved ade-quate only because the gene was largeand in a gene-poor region of the genome.

Subsequent successes with a series ofdisease genes reveal the influence of im-proved techniques. Examples in whichYACs, FISH, or both figured promi-nently include the following: fragile-Xsyndrome (48-50), the most commonheritable form of mental retardation; fa-milial adenomatous polyposis (51, 52), aheritable form of colorectal cancer; my-otonic dystrophy (53), an adult-onset dis-ease that affects muscle function; Kall-mann syndrome (54, 55), a defect in neu-ronal development; Lowe syndrome (56),a developmental defect affecting the lens,brain, and kidney; and Menkes disease

4340 Review: Olson

Proc. Natl. Acad. Sci. USA 90 (1993) 4341

(57-59), a neurological disease that islethal in early childhood.The genes for many other heritable

diseases are now under analysis by sim-ilar techniques. Particularly impressive isprogress on the genetic mapping of dis-eases such as familial breast and ovariancancer (60, 61) and early-onset familialAlzheimer disease (62). The genetic anal-ysis of these diseases is complicated by aset of factors that will be encounteredincreasingly often as positional cloning isapplied to complex, adult-onset geneticdisorders: suitable families are rare,small, and incomplete (i.e., few grand-parents, parents, or siblings of the af-fected individuals are available); evenfamily members that remain disease-freethroughout a normal life span cannot bereliably categorized as unaffected sincethey may have died from other causesbefore disease developed; the disease iscommon enough in the general popula-tion that cases with genetic and nonge-netic causes occur frequently in the samefamily. Highly informative genetic mark-ers, such as the PCR-detectable CA-repeat polymorphisms, have helped ad-dress these problems since they maxi-mize the likelihood that the segment ofthe chromosome that bears the disease-causing mutation can be tracked reliablyfrom one generation to the next evenwhen there are many family membersthat are missing or must be excluded fromthe study because of their uncertain dis-ease status.

In addition to improved genetic map-ping, successful completion of theseprojects may require further advances inphysical mapping and sequencing. Be-cause of the difficulty of the genetic anal-ysis, it is unlikely that disease genes suchas the recently described one for early-onset familial Alzheimer disease on chro-mosome 14 (62) will be localized even towithin 1 Mbp by genetic mapping. Thus,its isolation will place great demands onphysical mapping resources and tech-niques for locating genes within clonedDNA. The case of Huntington disease,for which ample family resources areavailable, is instructive: the gene wasgenetically mapped to a position near theend of the short arm of chromosome 4 in1983 (63) but has not yet been identifiedin cloned DNA. Its position, even now, isknown only to within 2.5 Mbp (64).

Still another class of disease geneswhose analysis has benefited from newinfrastructure are genes whose disruptionin somatic cells causes cancer. Particu-larly in leukemias and lymphomas, acommon mechanism by which disease-causing mutations arise is translocation,a process of chromosome breakage andrejoining. The combination ofYACs andFISH analysis has simplified the mappingof the chromosomal breakpoints and al-lowed the isolation of genes whose dis-

ruption is the initiating event in severalforms of neoplasia (65, 66).

DNA Sequence as an Interface BetweenKnowledge of Human Biology andKnowledge of the Biology ofModel Organisms

Central to the NRC committee's recom-mendations, which emphasized the im-portance of sequencing the genomes ofmodel organisms, was the belief thatDNA sequence offers a potent means ofinterrelating diverse aspects of biologicalknowledge. Events during the past 5years have strongly reinforced this con-cept.

Particularly remarkable is the ability ofDNA-sequence data to call attention tosimilarities between biological phenom-ena that are superficially unrelated. Atypical example involves the successfultransfer of information from the study ofyeast mating to diverse areas of humanbiology. Yeast cells have two matingtypes, commonly referred to as a and a.In the yeast life cycle, a and a cells arethe rough counterparts of mammaliangerm cells. The yeast counterpart of fer-tilization involves the fusion of a and acells, a process that is partly mediated bytwo peptide hormones, a factor and afactor. These hormones are named afterthe cell type that secretes them. Theytrigger a series of changes in the oppositecell type that prepare the cell for fusion.The mechanisms through which a fac-

tor and a factor are synthesized andsecreted have been studied in detail bygenetic techniques that are particularlywell developed in yeast (67). A peculiarfeature of a-factor secretion is its inde-pendence of the pathway through whichyeast proteins are normally secreted. Aparticular gene, STE6, encodes a proteinthat allows a factor to leave the cell whilebypassing the normal secretory pathway.Sequence analysis of STE6 revealed un-mistakable similarity to the human genemdrl (68, 69). This gene has attractedinterest because of its involvement inmultiple-drug resistance, a phenomenonin which malignant cells become simul-taneously resistant to several of the mostcommonly used chemotherapeuticagents (70). Once the relatedness ofSTE6and mdrl had been established by se-quence comparison, it was quicklyshown by gene-transfer experiments thatthe mouse version of the mdrl gene willactually substitute in yeast for the func-tion of STE6, correcting the inability ofyeast cells with mutations in the STE6gene to secrete a factor (71). The avail-ability of the yeast system opens up apowerful new front for the study of thispoorly understood transport mechanism.

Studies of a-factor biosynthesis haveproven equally productive in providinginsights into human metabolism. While

the secretion of a factor follows the nor-mal pathway, its biosynthesis has a fea-ture that is unusual in yeast but relativelycommon in human cells: it is produced byproteolytic processing of a precursorpeptide at a Lys-Arg linkage. Geneticstudies revealed that the gene KEX2 en-codes the protease that carries out thisprocessing step. Comparison of the se-quence of KEX2 with all other knownDNA sequences revealed strong similar-ity to a human gene of previously un-known function, c-fur (72). Subsequentanalysis showed that c-fur is a member ofa family of human genes that encodeproteases that process precursors tomany important proteins and peptidesincluding insulin, nerve growth factor,bone morphogenetic protein, and a majorcomponent of the AIDS virus (73). Thereis a long history of direct, biochemicalefforts to identify these proteases be-cause of their potential interest as phar-macological targets. These efforts led tothe description of a whole series of pro-teases that are capable of cleaving Lys-Arg linkages under particular in vitroconditions but that serve other functionsin vivo.These two examples illustrate the

strength of the concept, which is funda-mental to the Human Genome Project,that DNA sequence provides the key toefficient knowledge transfer betweenmodel organisms and human biology. Atpresent, this process requires consider-able serendipity, only because availableDNA-sequence data on the genomes ofboth the human and the major modelorganisms are fragmentary. There hasbeen enough progress on the sequencingof the genomes of E. coli (74), S. cerevi-siae (75), and the nematode worm C.elegans (76) to indicate the value of sys-tematic genomic sequencing. However,the real work of determining completesequences for the genomes of thesemodel organisms still lies ahead.

In humans, the main new source ofsystematic data has come from the se-quencing of cDNAs, cloned DNA copiesof the messenger RNA (mRNA) mole-cules that actually direct protein synthe-sis (ref. 4, pp. 102-104; ref. 77). Thismethod is a cost-effective way of discov-ering new human genes because only asmall fraction of genomic DNA directlycodes for proteins. However, cDNA se-quencing is unlikely to replace genomicsequencing as the definitive method ofcharacterizing the complete set ofhumangenes for several reasons: genes containcritical DNA sequences that regulatetheir expression but are not included inthe mRNA; there are common instancesboth in which one gene produces multi-ple, substantially different mRNA mole-cules and in which multiple genes pro-duce nearly identical mRNA molecules,situations that are difficult to sort out

Review: Olson

Proc. Natl. Acad. Sci. USA 90 (1993)

without detailed knowledge of the struc-tures of the corresponding genes andgene families; the cost advantages ofcDNA sequencing erode when the goal isaccurate, full-length sequences ratherthan one-pass, partial sequences; no ad-equate solution has been found to theproblem that the mRNA products of dif-ferent genes are present at widely differ-ent concentrations, which vary dramati-cally in different tissues, different devel-opmental stages, and different metabolicstates.

Open-Ended Improvements in theAnalytical Biochemistry of DNA

The promise of DNA-sequence compar-ison as a fundamental tool in biologicalresearch emphasizes the need for pro-gressively better methods of DNA anal-ysis, particularly DNA sequencing. Acritical feature of this challenge is itsopen-ended nature. DNA sequencing is atechnology, like digital computing, forwhich there is no obvious point at whichfurther improvements would saturate po-tential applications. A basic misimpres-sion about the Human Genome Project isthat once its narrow goals are met, de-mands for large-scale DNA sequencingwill taper off. DNA sequence data arebasically a source of hypotheses, therigorous testing of which typically re-quires the acquisition of still more DNAsequence. The determination of a "ref-erence" human sequence will provide astrong incentive to trace genes throughevolution with finer grain than the E.coli/yeast/worm/fly/mouse/humancomparisons on which the NRC commit-tee recommended early emphasis. Fi-nally, the study of individual variation,which plays a central role both in biologyand in medicine, poses unbounded de-mands for DNA-sequence data.Juxtaposed to this open-ended need for

improvements in the efficiency of DNAsequencing is the reality that there hasbeen no obvious increase in the basic effi-ciency ofDNA sequencing during the pastdecade. The protocols have become morerobust, and the skill level required forsuccess has been lowered. Fluorescence-based methods with real-time detection ofthe products ofDNA-sequencing reactionsduring electrophoresis have eased labora-tory management of large projects anddecreased the subjectivity ofdata interpre-tation (ref. 4, pp. 595-598; ref. 78). Hence,the practicality of large projects is greaternow than it was a decade ago. However, itis not apparent that there has been anychange in either the efficiency or the accu-racy with which an expertDNA sequencercan gather data.The NRC committee recognized this

problem but was overoptimistic about itsresolution (ref. 1, p. 2):

"The technical problems associatedwith mapping and sequencing the hu-man and other genomes are sufficientlygreat that a scientifically sound programrequires a diversified, sustained effortto improve our ability to analyze com-plex DNA molecules .... Prospectsare ... good that the required ad-vanced DNA technologies wouldemerge from a focused effort that em-phasizes pilot projects and technologi-cal development."

NIH and DOE have both made vigorousefforts to steer a significant portion of theproject in the recommended directions.However, there is little indication thatdecisive research momentum has devel-oped in the technology of DNA sequenc-ing. It can be argued that the problem ispredominantly cultural rather than tech-nical, relating to the different value sys-tems and research emphases of molecu-lar genetics, on the one hand, and ana-lytical chemistry, applied physics, andengineering, on the other (79).An illustration of the magnitude of the

technical challenge is provided by the gapbetween the theoretical and actual outputof the current generation of DNA-sequencing instruments. Standard com-mercial instruments now have the capac-ity to produce -30 kilobase pairs (kbp) ofraw sequence data per day (78). Allowingfor the desirability of determining thesequence of the two redundant strands ofDNA independently and for some over-sampling of data for each strand, a ratioof raw sequence data to finished data of5:1 should be achievable. Hence, a singleinstrument should be capable of produc-ing 6 kbp offinished sequence per day, or=2 Mbp per year. In reality, no genomecenter has yet produced even 1 Mbp ofcontiguous, finished sequence per yeareven though such centers typically havemany sequencing instruments.

This paradox reflects the present im-possibility of integrating all the steps inDNA sequencing into a continuous pro-cess that fully utilizes even the capabili-ties of current sequencing instruments.Although this experience is universalamong DNA sequencing laboratories,there is little consensus about whichsteps in the process are rate limiting,much less what should be done to im-prove them.What is clear is that there is a dramatic

gap between the advanced biologicaltechnologies of molecular genetics andthe primitive nonbiological technologies.The latter include the physical manipu-lation of samples, methods of chemicaland physical analysis, process design,quality control, and information han-dling. These areas are all critical to ef-forts to scale up bench-top moleculargenetics, and most biologists are poorlytrained to make the needed innovations.

The Human Genome Project has stimu-lated increased interactions between bi-ologists and scientists and technologistswho have the necessary expertise tosolve these problems. However, the dif-ficulties of translating these beginningsinto major improvements in DNA analy-sis continue to pose substantial policychallenges.

Conclusions

For an effort that is only in its third yearof substantial funding, the Human Ge-nome Project in the United States ismaking good progress toward its centralgoals. The policy on which it was basedhas proven farsighted even in the face ofrapid technological change. In the map-ping goals of the project, which havedominated the first years, the experimen-tal methods that are leading to successhave diverged widely from those extantwhen the NRC report was issued. None-theless, the report's conceptual frame-work has survived with little alteration.Examples abound of biological ad-

vances that have benefited directly fromthe early activities ofthe Human GenomeProject. Precise tracking of the cause-and-effect relationships between activi-ties funded through the Human GenomeProject in the United States and specificbiological advances is neither possiblenor desirable. Human genome analysis isa loosely coordinated international en-deavor to which funding agencies andscientists in many countries have alreadymade important contributions. Vigorousresearch activity funded through otherFederal programs, private agencies, andindustry has also had a major impact.Nonetheless, NIH and DOE program-matic efforts, particularly through theirproductive investment in YACs, FISH,PCR-detectable DNA polymorphisms,and radiation-hybrid mapping, haveclearly achieved good progress towardthe mapping goals of the NRC report andalso contributed directly to the success ofmany other research projects in the bio-medical sciences.

Like other human endeavors, the Hu-man Genome Project has succeeded bestwhen it has aligned itself with broadertrends. Examples include its increasingreliance on PCR, yeast genetics, and flu-orescence microscopy. It has succeededleast when it has tried to establish newtrends such as the importation of hightechnology from other areas into biology.This tension is healthy and will undoubt-edly remain as the project focuses in-creased attention on its flagship goal ofdetermining the sequence of the 990% ofthe human genome about which we stillknow almost nothing.

Note Added in Proof. The gene that is mutated

4342 Review: Olson

Proc. Natl. Acad. Sci. USA 90 (1993) 4343

in Huntington disease has now been identified(80).

1. National Research Council (1988) Map-ping and Sequencing the Human Ge-nome (Natl. Acad. Press, Washington,DC).

2. Smith, H. 0. & Wilcox, K. W. (1970) J.Mol. Biol. 51, 379-391.

3. Smith, H. 0. (1979) Science 205, 455-462.

4. Watson, J. D., Gilman, M., Witkowski,J. & Zoller, M. (1992) RecombinantDNA(Freeman, New York), 2nd Ed.

5. Nathans, D. (1979) Science 206, 903-909.6. Botstein, D., White, R. L., Skolnick, M.

& Davis, R. W. (1980) Am. J. Hum.Genet. 32, 314-331.

7. Southern, E. M. (1975) J. Mol. Biol. 98,503-517.

8. Donis-Keller, H., Green, P., Helms, C.,Cartinhour, S., Weiffenbach, B. et al.(1987) Cell 51, 319-337.

9. Sanger, F., Nicklen, S. & Coulson, A. R.(1977) Proc. Natl. Acad. Sci. USA 74,5463-5467.

10. Maxam, A. M. & Gilbert, W. (1977)Proc. Natl. Acad. Sci. USA 74, 560-564.

11. Baer, R., Bankier, A. T., Biggin, M. D.,Deininger, P. L., Farrell, P. J., Gibson,T. J., Hatfull, G., Hudson, G. S., Satch-well, S. C., Seguin, C., Tuffnell, P. S. &Barrell, B. G. (1984) Nature (London)310, 207-211.

12. Alberts, B., Bray, D., Lewis, J., Raff,M., Roberts, K. & Watson, J. D. (1989)Molecular Biology of the Cell (Garland,New York), 2nd Ed.

13. Fink, G. R. (1988) Genetics 118, 549-550.

14. Beaucage, S. L. & Caruthers, M. H.(1981) Tetrahedron Lett. 22, 1859-1862.

15. Caruthers, M. H. (1985) Science 230,281-285.

16. Hunkapiller, M., Kent, S., Caruthers,M., Dreyer, W., Firca, J., Giffin, C.,Horvath, S., Hunkapiller, T., Tempst, P.& Hood, L. (1984) Nature (London) 310,105-111.

17. Saiki, R. K., Scharf, S., Faloona, F.,Mullis, K. B., Horn, G. T., Erlich,H. A. & Arnheim, N. (1985) Science 230,1350-1354.

18. Saiki, R. K., Gelfand, D. H., Stoffel, S.,Scharf, S. J., Higuchi, R., Horn, G. T.,Mullis, K. B. & Erlich, H. A. (1988)Science 239, 487-491.

19. Olson, M., Hood, L., Cantor, C. & Bot-stein, D. (1989) Science 245, 1434-1435.

20. Coulson, A., Sulston, J., Brenner, S. &Karn, J. (1986) Proc. Natl. Acad. Sci.USA 83, 7821-7825.

21. Olson, M. V., Dutchik, J. E., Graham,M. Y., Brodeur, G. M., Helms, C.,Frank, M., MacColin, M., Scheinman,R. & Frank, T. (1986) Proc. Natl. Acad.Sci. USA 83, 7826-7830.

22. Kohara, Y., Akiyama, K. & Isono, K.(1987) Cell 50, 495-508.

23. Green, E. D. & Olson, M. V. (1990) Sci-ence 250, 94-98.

24. Burke, D. T., Carle, G. F. & Olson,M. V. (1987) Science 236, 806-812.

25. Coulson, A., Kozono, Y., Lutterbach,B., Shownkeen, R., Sulston, J. & Wa-terston, R. (1991) BioEssays 13, 413-417.

26. Brownstein, B. H., Silverman, G. A.,Little, R. D., Burke, D. T., Korsmeyer,S. J., Schlessinger, D. & Olson, M. V.(1989) Science 244, 1348-1351.

27. Anand, R., Ogilvie, D. J., Butler, R.,Riley, J. H., Finniear, R. S., Powell,S. J., Smith, J. C. & Markham, A. F.(1991) Genomics 9, 124-130.

28. Silverman, G. A., Jockel, J. I., Domer,P. H., Mohr, R. M., Taillon-Miller, P. &Korsmeyer, S. J. (1991) Genomics 9,219-228.

29. Little, R. D., Pilia, G., Johnson, S.,D'Urso, M. & Schlessinger, D. (1992)Proc. Natl. Acad. Sci. USA 89, 177-181.

30. Chumakov, I., Rigault, P., Guillou, S.,Ougen, P., Billaut, A. et al. (1992) Nature(London) 359, 380-387.

31. Foote, S., Vollrath, D., Hilton, A. &Page, D. C. (1992) Science 258, 60-66.

32. Gall, J. G. & Pardue, M. L. (1969) Proc.Natl. Acad. Sci. USA 63, 378-383.

33. Landegent, J. E., Jansen in de Wal, N.,van Ommen, G.-J. B., Baas, F., de Vi-jlder, J. J. M., van Duijn, P. & van derPloeg, M. (1985) Nature (London) 317,175-177.

34. Landegent, J. E., Jansen in de Wal, N.,Dirks, R. W., Baas, F. & van der Ploeg,M. (1987) Hum. Genet. 77, 366-370.

35. Lawrence, J. B., Villnave, C. A. &Singer, R. H. (1988) Cell 52, 51-61.

36. Trask, B., Pinkel, D. & van den Engh, G.(1989) Genomics 5, 710-717.

37. Lichter, P., Tang, C.-J. C., Call, K.,Hermanson, G., Evans, G. A., Hous-man, D. & Ward, D. (1990) Science 247,64-69.

38. Stallings, R. L., Torney, D. C., Hilde-brand, C. E., Longmire, J. L., Deaven,L. L., Jett, J. H., Doggett, N. A. &Moyzis, R. K. (1990) Proc. Natl. Acad.Sci. USA 87, 6218-6222.

39. Tynan, K., Olsen, A., Trask, B., deJong, P., Thompson, J., Zimmermann,W., Carrano, A. & Mohrenweiser, H.(1992) Nucleic Acids Res. 20, 1629-1636.

40. Cox, D. R., Burmeister, M., Price,E. R., Kim, S. & Myers, R. M. (1990)Science 250, 245-250.

41. Weber, J. L. & May, P. E. (1989) Am. J.Hum. Genet. 44, 388-396.

42. Litt, M. & Luty, J. A. (1989) Am. J.Hum. Genet. 44, 397-401.

43. NIH/CEPH Collaborative MappingGroup (1992) Science 258, 67-86.

44. Weissenbach, J., Gyapay, G., Dib, C.,Vignal, A., Morissette, J., Millasseau,P., Vaysseix, G. & Lathrop, M. (1992)Nature (London) 359, 794-801.

45. Dietrich, W., Katz, H., Lincoln, S. E.,Shin, H. S., Friedman, J., Dracopoli,N. C. & Lander, E. S. (1992) Genetics131, 423-447.

46. Collins, F. S. (1992) Nature Genet. 1,3-6.

47. Rommens, J. M., Iannuzzi, M. C.,Kerem, B. S., Drumm, M. L., Melmer,G., Dean, M., Rozmahel, R., Cole, J. L.,Kennedy, D., Hidaka, N., Zsiga, M.,Buchwald, M., Riordan, J. R., Tsui,L. C. & Collins, F. S. (1989) Science245, 1059-1065.

48. Verkerk, A. J. M. H., Pieretti, M., Sut-cliffe,J. S., Fu,Y. H., Kuhl, D. P. A. etal. (1991) Cell 65, 905-914.

49. Kremer, E. J., Pritchard, M., Lynch,

M., Yu, S., Holman, K., Baker, E.,Warren, S. T., Schlessinger, D., Suther-land, G. R. & Richards, R. I. (1991) Sci-ence 252, 1711-1714.

50. Oberle, I., Rousseau, F., Heitz, D.,Kretz, C., Devys, D., Hanauer, A.,Boue, J., Bertheas, M. F. & Mandel,J. L. (1991) Science 252, 1097-1102.

51. Kinzler, K. W., Nilbert, M. C., Su,L. K., Vogelstein, B., Bryan, T. M. etal. (1991) Science 253, 661-665.

52. Groden, J., Thliveris, A., Samowitz, W.,Carlson, M., Gelbert, L. et al. (1991) Cell66, 589-600.

53. Fu, Y. H., Pizzuti, A., Fenwick, R. G.,Jr., King, J., Rajnarayan, S., Dunne,P. W., Dubel, J., Nasser, G. A., Ash-izawa, T., de Jong, P., Wieringa, B.,Korneluk, R., Perryman, M. B., Ep-stein, H. F. & Caskey, C. T. (1992) Sci-ence 255, 1256-1258.

54. Legouis, R., Hardelin, J. P., Levilliers,J., Claverie, J. M., Compain, S.,Wunderle, V., Millasseau, P., Le Paslier,D., Cohen, D., Caterina, D., Bouguel-eret, L., Delemarre-Van de Waal, H.,Lutfalla, G., Weissenbach, J. & Petit, C.(1991) Cell 67, 423-435.

55. Franco, B., Guioli, S., Pragliola, A.,Incerti, B., Bardoni, B., Tonlorenzi, R.,Carrozzo, R., Maestrini, E., Pieretti, M.,Taillon-Miller, P., Brown, C. J., Willard,H. F., Lawrence, C., Persico, M. G.,Camerino, G. & Ballabio, A. (1991) Na-ture (London) 353, 529-536.

56. Attree, O., Olvios, I. M., Okabe, I., Bai-ley, L., Nelson, D. L., Lewis, R. A.,McInnes, R. R. & Nussbaum, R. L.(1992) Nature (London) 358, 239-242.

57. Vulpe, C., Levinson, B., Whitney, S.,Packman, S. & Gitschier, J. (1993) Na-ture Genet. 3, 7-13.

58. Chelly, J., Tumer, Z., Tonnesen, T.,Petterson, A., Ishikawa-Brush, Y., Tom-merup, N., Horn, N. & Monaco, A. P.(1993) Nature Genet. 3, 14-19.

59. Mercer, J. F. B., Livingston, J., Hall,B., Paynter, J. A., Begy, C., Chan-drasekharappa, S., Lockhart, P.,Grimes, A., Bhave, M., Siemieniak, D.& Glover, T. W. (1993) Nature Genet. 3,20-25.

60. Hall, J. M., Lee, M. K., Newman, B.,Morrow, J. E., Anderson, L. A., Huey,B. & King, M.-C. (1990) Science 250,1684-1689.

61. Hall, J. M., Friedman, L., Guenther, C.,Lee, M. K., Weber, J. L., Black, D. M.& King, M.-C. (1992) Am. J. Hum.Genet. 50, 1235-1242.

62. Schellenberg, G. D., Bird, T. D., Wijs-man, E. M., Orr, H. T., Anderson, L.,Nemens, E., White, J. A., Bonnycastle,L., Weber, J. L., Elisa Alonso, M., Pot-ter, H., Heston, L. L. & Martin, G. M.(1992) Science 258, 668-671.

63. Gusella, J. F., Wexler, N. S., Con-neally, P. M., Naylor, S. L., Anderson,M. A., Tanzi, R. E., Watkins, P. C., Ot-tina, K., Wallace, M. R., Sakaguchi,A. Y., Young, A. B., Shoulson, I., Be-nilla, E. & Martin, J. B. (1983) Nature(London) 306, 234-238.

64. Davies, K. (1992) Nature (London) 357,page before 95.

65. Diabali, M., Selleri, L., Parry, P.,Bower, M., Young, B. D. & Evans,

Review: Olson

4344 Review: Olson Proc. Natl. Acad. Sci. USA 90 (1993)

G. A. (1992) Nature Genet. 2, 113-118.

66. Ziemin-van der Poel, S., McCabe, N. R.,Gill, H. J., Espinosa, R., III, Patel, Y.,Harden, A., Rubinelli, P., Smith, S. D.,LeBeau, M. M., Rowley, J. D. & Diaz,M. 0. (1991) Proc. Nati. Acad. Sci. USA88, 10735-10739.

67. Botstein, D. & Fink, G. R. (1988) Sci-ence 240, 1439-1443.

68. Kuchler, K., Sterne, R. E. & Thorner, J.(1989) EMBO J. 8, 3973-3984.

69. McGrath, J. P. & Varshavsky, A. (1989)Nature (London) 340, 400-404.

70. Gottesman, M. M. & Pastan, I. (1988) J.Biol. Chem. 263, 12163-12166.

71. Raymond, M., Gros, P., Whiteway, M.

& Thomas, D. Y. (1992) Science 256,232-234.

72. Fuller, R. S., Brake, A. J. & Thorner, J.(1989) Science 246, 482-486.

73. Barr, P. J. (1991) Cell 66, 1-3.74. Daniels, D. L., Plunkett, G., III, Bur-

land, V. & Blattner, F. R. (1992) Science257, 771-778.

75. Oliver, S. G., van der Aart, Q. J. M.,Agostoni-Carbone, M. L., Aigle, M., Al-berghina, L. et al. (1992) Nature (Lon-don) 357, 38-46.

76. Sulston, J., Du, Z., Thomas, K., Wilson,R., Hillier, L., Staden, R., Halloran, N.,Green, P., Thierry-Mieg, J., Qiu, L.,Dear, S., Coulson, A., Craxton, M.,Durbin, R., Berks, M., Metzstein, M.,

Hawkins, T., Ainscough, R. & Water-ston, R. (1992) Nature (London) 356,37-41.

77. Adams, M. D., Kelley, J. M., Gocayne,J. D., Dubnick, M., Polymeropoulos,M. H., Xiao, H., Merril, C. R., Wu, A.,Olde, B., Moreno, R. F., Kerlavage,A. R., McCombie, W. R. & Venter,J. C. (1991) Science 252, 1651-1656.

78. Hunkapiller, T., Kaiser, R. J., Koop,B. F. & Hood, L. (1991) Science 254,59-67.

79. Olson, M. V. (1991) Anal. Chem. 63,416A-420A.

80. The Huntington's Disease CollaborativeResearch Group (1993) Cell 72, 971-983.