The past, present and future of Scientific discourse.

The science journal is 346 years old in 2011, having evolved continuously but largely incrementally over that period. Its reinvention for an online presence has largely preserved its previously printed nature, in the sense that much of the increased functionality which is potentially offered by this new medium has yet to be exploited. In the present article an attempt is made to discuss two previously published papers, one in 1953 and the other in 2010, and to illustrate how additional functionality can be implemented in the form of accessible data sourced from quantum mechanical calculation and how subsequent discourse in the form of blogs may add to the process. In this sense, the reader of this article is invited to try for themselves whether these enhancements improve their scientific understanding, and whether such enhanced journals are good models for the future evolution of the genre.

Introduction

The first journal devoted exclusively to science is generally accepted to have first appeared in 1665 as the Philosophical Transactions (of the Royal Society). The inaugural issue [1], which famously carries an account by Robert Boyle of "a very odd monstrous n class="Species">calf", is perhaps class="Chemical">not scieclass="Chemical">nce as we kclass="Chemical">now it class="Chemical">nowadays, but it does remiclass="Chemical">nd oclass="Chemical">ne rather of what oclass="Chemical">ne might ficlass="Chemical">nd iclass="Chemical">n a persoclass="Chemical">nal blog, a similarity I will returclass="Chemical">n to later. The structure of the scieclass="Chemical">ntific jourclass="Chemical">nal aclass="Chemical">nd the articles published by this meaclass="Chemical">ns evolved coclass="Chemical">nstaclass="Chemical">ntly aclass="Chemical">nd mostly iclass="Chemical">ncremeclass="Chemical">ntally duriclass="Chemical">ng the class="Chemical">next 330 years. Oclass="Chemical">ne of the more sigclass="Chemical">nificaclass="Chemical">nt, but class="Chemical">nevertheless still iclass="Chemical">ncremeclass="Chemical">ntal chaclass="Chemical">nges was addiclass="Chemical">ng aclass="Chemical">n oclass="Chemical">nliclass="Chemical">ne preseclass="Chemical">nce duriclass="Chemical">ng the late 1990s. Iclass="Chemical">ndeed, some jourclass="Chemical">nals fouclass="Chemical">nded iclass="Chemical">n the late 1990s offered oclass="Chemical">nly aclass="Chemical">n oclass="Chemical">nliclass="Chemical">ne versioclass="Chemical">n [2] aclass="Chemical">nd there are sigclass="Chemical">ns that some older jourclass="Chemical">nals may be prepariclass="Chemical">ng to abaclass="Chemical">ndoclass="Chemical">n the (relatively expeclass="Chemical">nsive) priclass="Chemical">nted form. Access to the oclass="Chemical">nliclass="Chemical">ne versioclass="Chemical">ns iclass="Chemical">n 2011 is predomiclass="Chemical">naclass="Chemical">ntly via a format which is perhaps best described as digital paper (PDF), although most jourclass="Chemical">nals also offer the articles iclass="Chemical">n aclass="Chemical">n alterclass="Chemical">native hypertext (HTML) format. There is class="Chemical">no jourclass="Chemical">nal yet that has adopted a format iclass="Chemical">ncreasiclass="Chemical">ngly used for books, the epub/epub3 staclass="Chemical">ndard [3]. Oclass="Chemical">n the horizoclass="Chemical">n is also the fifth major evolutioclass="Chemical">n of hypertext markup laclass="Chemical">nguage kclass="Chemical">nowclass="Chemical">n as HTML5 [4,5], which strives to offer a richer iclass="Chemical">nteractive medium to the reader. Certaiclass="Chemical">nly machiclass="Chemical">nes (e.g. such as those workiclass="Chemical">ng oclass="Chemical">n behalf of search eclass="Chemical">ngiclass="Chemical">nes such as Google) class="Chemical">now also automatically process the articles published iclass="Chemical">n the moderclass="Chemical">n jourclass="Chemical">nal, iclass="Chemical">ndexiclass="Chemical">ng the full-textual coclass="Chemical">nteclass="Chemical">nt, addiclass="Chemical">ng rich metadata oclass="Chemical">n the topics thereiclass="Chemical">n described, aclass="Chemical">nd class="Chemical">noticlass="Chemical">ng (but largely iclass="Chemical">ncapable of truly iclass="Chemical">ndexiclass="Chemical">ng the coclass="Chemical">ntext of) the images aclass="Chemical">nd figures. Software caclass="Chemical">n also usefully replace the old processes of biclass="Chemical">ndiclass="Chemical">ng the priclass="Chemical">nted jourclass="Chemical">nal aclass="Chemical">nd the storage of volumes oclass="Chemical">n shelves iclass="Chemical">n a library or aclass="Chemical">n office, a process delightfully described (by biologists, class="Chemical">not chemists) as defrosticlass="Chemical">ng the digital library[6]. These processes largely address oclass="Chemical">nly the bibliographic issues (via rich metadata harvesticlass="Chemical">ng) rather thaclass="Chemical">n attempticlass="Chemical">ng to defrost the scieclass="Chemical">ntific or chemical coclass="Chemical">nteclass="Chemical">nt itself. It is the issues iclass="Chemical">nvolved iclass="Chemical">n defrosticlass="Chemical">ng the latter type of iclass="Chemical">nformatioclass="Chemical">n aclass="Chemical">nd data that the preseclass="Chemical">nt article addresses.

Data has always been the "n class="Species">elephant iclass="Chemical">n the room" of scieclass="Chemical">ntific publishiclass="Chemical">ng. Because the costs of priclass="Chemical">nticlass="Chemical">ng aclass="Chemical">nd distributiclass="Chemical">ng paper are still sigclass="Chemical">nificaclass="Chemical">nt to this day, priclass="Chemical">nt was class="Chemical">never really beeclass="Chemical">n coclass="Chemical">nsidered a viable mechaclass="Chemical">nism for distributiclass="Chemical">ng the (ofteclass="Chemical">n very large amouclass="Chemical">nts of) data, whclass="Chemical">n class="Chemical">ether raw, or partially processed, on which almost all scientific models, theories and their interpretations are based. Instead, starting in the early 1990s and coincident with the first introduction of the Internet, many science journals offered an annex to the main journal in the form of supporting or supplemental information. This was provided in final form by the authors themselves, and the journal itself added little extra value such as indexing to this (often purely visual) content. It was very much up to an interested reader to add their own value to any (visual, textual or numerical) supporting data that might be associated with an article.

The long view over a 350 year period is that these evolutions of the journal could be regarded as largely relating to the production and delivery processes of journals, and arguably have not been matched by similar advances in how scientists consume or use journals. In this essay, I will analyze two chemical articles, published in respectively 1953 and 2010, from the point of view of how the original journal presented the scientific discourse, what the limitations of that presentation might have been, and the prospects of how it could evolve into a step-change rather than incremental change in that discourse.

The relationship between a journal article and data

I start the analysis with article that contains (inter alia) what has been described as the most famous scientific diagram of the 20th Century, the representation [7] of the double helical structure of the DNA molecule by Watson and Crick (Figure 1).

Figure 1

The DNA double helix (reproduced with permission [8]), showing a right handed or B-helix.

The Dn class="Chemical">NA double helix (reproduced with permissioclass="Chemical">n [8]), showiclass="Chemical">ng a right haclass="Chemical">nded or B-helix.

Indeed, this diagram is the only one that actually appears in the article, and one would seek in vain any diagrammatic elaboration of what the molecular structure of DNA is (although components such as n class="Chemical">deoxyribose or class="Chemical">n class="Chemical">guanine are named as such in the text). Anyone seeking to repeat Watson and Crick's model building would certainly have to acquire additional molecular data from another sources. Some of that missing information is shown here in Figure 2, although this only describes the connectivity of the various atoms in a single strand of DNA, and not the two or three dimensional relationships of the (125 in this example) individual atoms. Note also that this diagram is presented here for visual consumption by a human, who still has to recover additional semantics such as the stereochemistry at the three stereogenic ribose centres, and note carefully that the unit represented must be accompanied by positively charged counter-ions.

Figure 2

The molecular basis of one strand of DNA, based on the CG bases.

The molecular basis of one strand of DNA, based on the n class="Chemical">CG bases.

Armed only with the one diagram actually published, curiosity might lead one to pose a scientific question such as "How did Watson and Crick assign the helix as right rather than left handed"? In other n class="Disease">words, oclass="Chemical">n what data did they base that coclass="Chemical">nclusioclass="Chemical">n? This does matter! For example, some 733 articles have appeared iclass="Chemical">n the scieclass="Chemical">nce literature over the last 20 years or so where DNA is represeclass="Chemical">nted as haviclass="Chemical">ng left-haclass="Chemical">nded helicity, iclass="Chemical">n most cases certaiclass="Chemical">nly erroclass="Chemical">neously [8]. Coiclass="Chemical">ncideclass="Chemical">ntally, similar issues of left or class="Chemical">n class="Disease">right-handedness were to be found when Pauling presented his α-helix models of proteins. In fact, almost all protein helices exhibit right-handedness [9]. A partial answer to that question is actually given in what is called the full version [10] of the preliminary article [7] (published as it happens by the Royal Society). Here we are told the following:

• that both chains follow right handed helices ...

• because left handed helices can only be constructed by violating permissible van der Waals contacts.

• We are informed that such permissible contacts include the approach of any two n class="Chemical">hydrogen atoms iclass="Chemical">n the molecule to a distaclass="Chemical">nce of class="Chemical">no less thaclass="Chemical">n 2.1Å.

• We are not however informed what the violations might be in a left handed helix that excludes this model. In other n class="Disease">words, just how close caclass="Chemical">n two class="Chemical">n class="Chemical">hydrogen atoms separated (for intramolecular contacts) by at least four bonds approach? In fact, distances of ~1.85Å or less have been observed [11].

In this same full article by Watson and Crick [10], we are given a table of numerical (polar) coordinates describing the positions of twelve key atoms, but it would have taken a very determined scientist to have used only this combination of information to easily confirm the assertion that a left-handed helix is excluded. Perhaps the lack of a model with which the reader could experiment might account for the relatively slow recognition of the importance of this article in the immediate years following its publication, and the observation that whilst a physical model of DNA had of course been built, it was only available for viewing (but not modifying) by visiting Cambridge!

One tool that modern chemistry now has at its disposal (which Watson and Crick did not have) are accurate molecular models based on quantum mechanical calculations. Such a molecule is quite a challenge to model, since the computation has to take into account subtle interactions such as dispersion (long range correlation) effects, which are more or less equivalent to the van der Waals contacts referred to by Watson and Crick, the ionic n class="Chemical">phosphate groups, the placlass="Chemical">nar bases aclass="Chemical">nd how they stack, so-called aclass="Chemical">nomeric effects at the base-class="Chemical">n class="Chemical">sugar connecting C-N bond, hydrogen bonds between both the obvious NH...N and NH...O atoms and less obvious ones such as C-H...O, and not least the capacity to deal self-consistently and accurately with the optimal positions of (at least) 250-254 atoms. In reality, such models have only very recently become available [12]. To illustrate how this famous article from 1953 [2] could now be published in a journal in 2011, I have taken the liberty of updating the original diagram with the one shown in Figure 3 (see additional file 1 for enhanced version) [6]. The additional information is made available via the figure caption and in Table 1 to conform to established practice in more conventional articles.

Figure 3

A model of the Z-d(CGCG). (a) Load coordinates for Z-d(CGCG)2 and (b) measure for close van der Waals contacts or (c) O...C contacts. (d) Load coordinates for the diastereomeric B-d(CGCG)2 and (e) view the O...H-N and C-H...O close contacts. (f) Load Z-d(ATAT)2 and (g) view the close O...C contacts. (h) Load B-d(ATAT)2 and (i) view the close O...H-N contacts.

Table 1

Relative thermodynamic energies (kcal mol-1)a

System	Total energy (duplex)	Dispersion contribution	ΔΔH298	Δ(-T. ΔS298)	ΔΔG298 duplex	ΔG298 single chain	ΔΔG298b
Z-CGCG	0.0	0.0	0.0	0.0	0.0	0.0	-60.3
B-CGCG	6.2	-5.1	8.0	+3.9	+11.9	+3.1	-54.7
Z-ATAT	0.0	0.0	0.0	0.0	0.0	0.0	-44.9
B-ATAT	-7.6	-12.5	-7.0	+2.7	-4.3	-1.8	-45.7

A model of the Z-d(n class="Chemical">CGCG). (a) Load coordiclass="Chemical">nates for Z-d(class="Chemical">n class="Chemical">CGCG)2 and (b) measure for close van der Waals contacts or (c) O...C contacts. (d) Load coordinates for the diastereomeric B-d(CGCG)2 and (e) view the O...H-N and C-H...O close contacts. (f) Load Z-d(ATAT)2 and (g) view the close O...C contacts. (h) Load B-d(ATAT)2 and (i) view the close O...H-N contacts.

This is a model of a Dn class="Chemical">NA duplex tetramer, built usiclass="Chemical">ng oclass="Chemical">nly the bases class="Chemical">n class="Chemical">CGCG or ATAT in this example, with inclusion of three phosphate groups and calculated for both the left- and right-handed helical form. The first of these was the one deprecated by Watson and Crick on the basis that the model violates permissible van der Waals contacts. The geometry is optimized to high convergence using a recent density functional formalism (ωB97XD) [13] which incorporates a correction for the attractive dispersion component of the van der Waals interactions. Justification for the use of this functional in describing hydrogen bonding has recently been published [14]. A 6-311G(d,p) basis set results in the wavefunction being described by up to 3468 basis functions, close to the practical limit using standard computing resources available in 2010. The ionic nature of the system, deriving from the phosphate groups, was treated using a self-consistent-reaction-field continuum solvent (water) [15,16] as implemented in the Gaussian09 package, revisions A.02 and B.01. 12. In such a model, duplex formation by combining two tri-anionic chains is nevertheless exothermic in the computed free energy (Table 1), which suggests the model is not physically unrealistic. Although in principle a full ion-pair resulting from inclusion of a solvated positive counterion (typically Na+ or NH4+ with additional water molecules) could also be treated using this method [17], the resulting model is too large and complex for the current available computational resources.

The resulting model is presented in this article using suitable software (Jmol in this instance [18]) which itself reads the optimized coordinates of all 250-254 atoms and renders these in suitable form for the reader. Annotation with identified close contacts between pairs of n class="Chemical">hydrogen atoms or other close coclass="Chemical">ntacts caclass="Chemical">n be easily scripted iclass="Chemical">n, aclass="Chemical">nd iclass="Chemical">n priclass="Chemical">nciple a rich variety of actioclass="Chemical">ns aclass="Chemical">nd aclass="Chemical">nalyses caclass="Chemical">n be built iclass="Chemical">nto the figure which are all based oclass="Chemical">n a combiclass="Chemical">natioclass="Chemical">n of the uclass="Chemical">nderlyiclass="Chemical">ng data aclass="Chemical">nd algorithms implemeclass="Chemical">nted by the (Jmol) software. Importaclass="Chemical">ntly, the origiclass="Chemical">nal data used for geclass="Chemical">neraticlass="Chemical">ng the model caclass="Chemical">n be extracted from the model (the process is described here [19]) aclass="Chemical">nd caclass="Chemical">n theclass="Chemical">n be re-applied usiclass="Chemical">ng alterclass="Chemical">native software which might provide further aclass="Chemical">nalysis, or iclass="Chemical">ndeed alterclass="Chemical">native techclass="Chemical">nologies such as stereoscopic processiclass="Chemical">ng. These processes class="Chemical">now turclass="Chemical">n the jourclass="Chemical">nal from merely a visual iclass="Chemical">nformatioclass="Chemical">n source iclass="Chemical">nto aclass="Chemical">n active scieclass="Chemical">ntific iclass="Chemical">nstrumeclass="Chemical">nt. We may also speculate at this poiclass="Chemical">nt oclass="Chemical">n other forms of reclass="Chemical">ndericlass="Chemical">ng data. Jmol was writteclass="Chemical">n iclass="Chemical">n Java, aclass="Chemical">nd requires the browser to support a Java virtual eclass="Chemical">nviroclass="Chemical">nmeclass="Chemical">nt. New geclass="Chemical">neratioclass="Chemical">ns of mobile iclass="Chemical">nformatioclass="Chemical">n devices, which are primarily desigclass="Chemical">ned for loclass="Chemical">ng battery life, may class="Chemical">not coclass="Chemical">nticlass="Chemical">nue with this approach. Iclass="Chemical">nstead, oclass="Chemical">ne favoured alterclass="Chemical">native is to iclass="Chemical">nterface the browser directly to the graphical hardware usiclass="Chemical">ng e.g. WebGL, aclass="Chemical">nd to implemeclass="Chemical">nt the fuclass="Chemical">nctioclass="Chemical">nality of somethiclass="Chemical">ng like Jmol usiclass="Chemical">ng the emergiclass="Chemical">ng HTML5 staclass="Chemical">ndard aclass="Chemical">nd appropriate scripts [20,21]. The class="Chemical">native ability of a browser to provide such eclass="Chemical">nhaclass="Chemical">nced processiclass="Chemical">ng is already appareclass="Chemical">nt iclass="Chemical">n support for SVG, a markup laclass="Chemical">nguage for vector graphics (examples of which are iclass="Chemical">ncluded iclass="Chemical">n Figure 4 [see additioclass="Chemical">nal file 2 for eclass="Chemical">nhaclass="Chemical">nced versioclass="Chemical">n]).

Figure 4

Calculated chiro-optical properties for DNA tetramers. aComputed at geometries optimised at the ωB97XD/6-31G(d) level with with application of a SCRF solvent continuum field for water. Chiro-optical properties computed at the CAM-B3LYP/6-31G(d,p) level with application of a SCRF solvent continuum field for water. ECD spectra computed at the TD-DFT level, using Nstates = 25 and a linewidth of 0.14 with application of a SCRF solvent continuum field for water. Click on image to expand the view of the ECD spectrum. Click on expanded view of spectrum to access the digital repository entry for that spectrum. bECD spectra are presented as scalable-vector-graphical diagrams (SVG). To view, use an SVG-capable browser.

Calculated chiro-optical properties for Dn class="Chemical">NA tetramers. aComputed at geometries optimised at the ωB97XD/6-31G(d) level with with applicatioclass="Chemical">n of a SCRF solveclass="Chemical">nt coclass="Chemical">nticlass="Chemical">nuum field for class="Chemical">n class="Chemical">water. Chiro-optical properties computed at the CAM-B3LYP/6-31G(d,p) level with application of a SCRF solvent continuum field for water. ECD spectra computed at the TD-DFT level, using Nstates = 25 and a linewidth of 0.14 with application of a SCRF solvent continuum field for water. Click on image to expand the view of the ECD spectrum. Click on expanded view of spectrum to access the digital repository entry for that spectrum. bECD spectra are presented as scalable-vector-graphical diagrams (SVG). To view, use an SVG-capable browser.

The reader is invited to load the Z-d(n class="Chemical">CGCG)2 coordiclass="Chemical">nates iclass="Chemical">n Figure 3. The leclass="Chemical">ngths of the class="Chemical">n class="Disease">van der Waals H...H attractions and the nucleophilic O...C attractions from the ribose ether oxygen to the electrophilic carbons on the guanine base have been enumerated. Because of the complex 3D nature of this molecule, they can only be truly perceived if the structure itself can be viewed from any desired angle (something clearly not possible in a conventional journal diagram). It also allows successive layers to be viewed, each perhaps concentrating on a particular aspect, without destroying or overwhelming the initial simple elegance of the overall concept (of a double helix). The identified ~2.8Å O...C interactions to the guanine are unique to the Z or left handed helical form, and since the sum of the vdW radii [22] of these two contact atoms is ~3.22Å, these are presumed to be (electrostatically) attractive. The alternative B-d(CGCG) 2 stereoisomer reveals these O...C contacts are absent, being instead replaced by hydrogen bonds (~1.9-2.1Å) between the ribose ether-oxygen and the NH2 hydrogens of the guanine. The sum of the O and H vdW radii is ~2.64Å, which suggests these are significantly attractive hydrogen bonds. There are additional C-H...O-P hydrogen bonded contacts of ~2.1-2.2Å. A similar divergence of attractive interactions emerges for chains built of AT bases. The Z-d(ATAT)2 duplex has only one ~2.8Å O...C interaction to the adenine, with three others having rather longer lengths (~3.0-3.1Å). The B-d(ATAT)2 duplex instead displays C-H...O contacts of ~2.4-2.5Å.

These differences can be more succinctly summarized as:

1. Z-d(n class="Chemical">CGCG)2 is stabilized (iclass="Chemical">nter alia) by a short coclass="Chemical">ntact betweeclass="Chemical">n a class="Chemical">n class="Chemical">carbon on the guanine and the ribose ether oxygen, of which there are four per four base pairs.

2. These contacts are replaced in B-d(n class="Chemical">CGCG)2 by NH class="Chemical">n class="Chemical">hydrogen bonded contacts to the ribose ether oxygen.

3. In Z-d(n class="Chemical">ATAT)2, the O-coclass="Chemical">ntacts to the class="Chemical">n class="Chemical">adenine are much longer, which

4. in B-d(n class="Chemical">ATAT)2, are replaced by class="Chemical">n class="Disease">short CH...O contacts.

By embedding access to accurate coordinate data within Figure 3, the reader can select whatever level of detail they desire from the diagram. Part of the origins of the relative stability the Z- and B- helical forms is not simply due to the presence (or in this case absence) of "violation of permissible van der Waals contacts", but also to several types of less common but nevertheless attractive interactions which may not have been inferred by building physical models alone. Such additional insights may in turn impact upon e. g. modeling one remarkable property of the DNA n class="Chemical">polymer, its ability to be stretched to almost twice its class="Chemical">normal leclass="Chemical">ngth without breakiclass="Chemical">ng [23].

It is of course the accumulation of these effects that determines the overall stability of the structure (Table 1). The thermodynamic quantities are computed with inclusion of thermal energies, obtained by solving the appropriate partition functions using calculated vibrational frequencies. Since these require second derivatives of energy with respect to coordinates, a smaller basis set 6-31G(d) was used for the purpose (a calculation time of ~4 days on a 12-core processor is typical). The dispersion corrections were obtained at the slightly higher 6-311G(d,p) basis set level to allow interactions to H to be modelled more realistically.

These energies reveal some surprises. Firstly, the free energy for forming the duplex from the separated chains is significantly exothermic, despite the electrostatic repulsions resulting from each chain carrying a 3- charge. For the resulting helix, the B-d(n class="Chemical">CGCG)2 form is 11.9 kcal/mol less stable iclass="Chemical">n terms of total free eclass="Chemical">nergy thaclass="Chemical">n the Z-isomer, but is 4.2 kcal/mol more stable for the dispersioclass="Chemical">n/vaclass="Chemical">n der Waals term, the criterioclass="Chemical">n suggested by Watsoclass="Chemical">n aclass="Chemical">nd Crick as geclass="Chemical">nerally discrimiclass="Chemical">naticlass="Chemical">ng agaiclass="Chemical">nst the Z-form (although without a speclass="Chemical">n class="Chemical">cification of the base type used for the model). The greater stability of the Z-form arises from a contribution of 3.9 from the entropy and 8.0 kcal/mol from the (zero-point energy corrected) enthalpy, which dominates the less favourable dispersion term.

The formation of a B-d(n class="Chemical">ATAT)2 duplex is less exothermic thaclass="Chemical">n that of the class="Chemical">n class="Chemical">CG duplex. It is now favoured by 5.2 kcal/mol over the Z-isomer in terms of free energy and by 12.8 kcal/mol in terms of dispersion contributions. The assertion often made [24] that the Z-helix is favoured by CG rich oligomers and the B-helix by AT-rich forms is thus confirmed by these calculations.

aThermochemistry computed at geometries optimised at the ωB97XD/6-31G(d) level with application of a SCRF solvent continuum field for n class="Chemical">water, with thermal correctioclass="Chemical">ns derived from computed vibratioclass="Chemical">nal frequeclass="Chemical">ncies. The dispersioclass="Chemical">n correctioclass="Chemical">ns are computed for geometries optimized at the ωB97XD/6-311G(d, p) level with applicatioclass="Chemical">n of a SCRF solveclass="Chemical">nt coclass="Chemical">nticlass="Chemical">nuum field for class="Chemical">n class="Chemical">water. The display coordinates are those obtained at this level. aFree energy for the dimerisation of a single strand to a duplex.

Armed with optimized coordinates which include the weaker interactions between atoms one can annotate the basic models revealed in Figure 3 with other (computed) properties. For example the n class="Disease">optical rotation [α]589 has the value +62° for Z-d(class="Chemical">n class="Chemical">CGCG)2 and -137° for the B-diastereomer, perhaps surprisingly small values for such an apparently asymmetric molecule. Also surprisingly, the corresponding experimental measurement does not appear to have been reported. Optical rotations are known to be rather fragile, being sensitive to small variations in conformation and solvation, but the electronic circular dichroism spectrum is regarded as rather more robust. Unlike other forms of spectroscopy such as NMR or IR, which can be used to infer structure from simple rules based on the functional groups present (in other words, local properties), these chiro-optical properties tend to be more characteristic of the global features of the molecule. As a result, they can be very difficult to interpret without a reasonably accurate model based on these global properties. A quantum mechanical computation of the molecular wavefunction is one such model, and it is now increasingly routinely used to help interpret optical rotations, electronic (and vibrational) circular dichroism spectra. Theory can now handle molecules containing ~250-254 atoms, such as the DNA tetramers modelled here. Annotation of the models with the calculated ECD spectra is included here in the hope they might prove useful for the interpretation of the experimental spectra.

This example has illustrated how access to accurate data can help provide additional insights into the factors controlling the stability of molecular structures. In this case, the factors controlling the helical stability of DNA duplexes can be teased out. By incorporating these models directly into the journal article (and providing links to digital repositories where a more complete dataset can be acquired if needed) the readers of the journal have an opportunity to discover their own insights within their own spheres of interest.

The crystal structure of 1,3-dimethylcyclobutadiene

The second example chosen for discussion is a more contemporary one. In July 2010, a article appeared [25] reporting the single-crystal X-ray structure of n class="Chemical">1,3-dimethylcyclobutadiene achieved by coclass="Chemical">nficlass="Chemical">nemeclass="Chemical">nt iclass="Chemical">n a crystalliclass="Chemical">ne matrix (Figure 5 [see additioclass="Chemical">nal file 3 for eclass="Chemical">nhaclass="Chemical">nced versioclass="Chemical">n]). The topic caught the imagiclass="Chemical">natioclass="Chemical">n, siclass="Chemical">nce class="Chemical">n class="Chemical">cyclobutadiene has been described as the Mona Lisa of molecules, and its very instability means that conventional experiments on it are very challenging. The article was however conventional in the sense of being made available in (more or less equivalent) HTML and PDF versions.

Figure 5

The reaction leading to 1,3-dimethylcyclobutadiene [25]. The numbering shown for 4 corresponds to that for the published coordinates. Load coordinates for host-guest structure and just the guest only.

The reaction leading to n class="Chemical">1,3-dimethylcyclobutadiene [25]. The class="Chemical">numbericlass="Chemical">ng showclass="Chemical">n for 4 correspoclass="Chemical">nds to that for the published coordiclass="Chemical">nates. Load coordiclass="Chemical">nates for host-guest structure aclass="Chemical">nd just the guest oclass="Chemical">nly.

Much of the scientific insight was carried in the form of four colour figures, all presented conventionally as single layered graphics with the viewpoint selected by the authors. Information on acquisition of the data on which these figures were based was given as citation 28 in that article, which lists deposition numbers (CCDC 764864-764868) and a URL that would enable a n class="Chemical">CIF file for each eclass="Chemical">ntry to be dowclass="Chemical">nloaded. Whilst this retrieval process is class="Chemical">not eclass="Chemical">ntirely automatic, it does take oclass="Chemical">nly a few miclass="Chemical">nutes to acquire the data. The usefulclass="Chemical">ness of the file is of course predicated oclass="Chemical">n the reader also haviclass="Chemical">ng access to appropriate software for aclass="Chemical">nalysis of a file iclass="Chemical">n this format. It is also importaclass="Chemical">nt to class="Chemical">note that a class="Chemical">n class="Chemical">CIF file allows inspection only of the refined crystallographic model presented in the article and the statistics associated with that model; it does not allow the user access to the underlying (hkl) diffraction data which would allow other models to be refined and assessed.

The reaction scheme reported [25] for photochemical generation of trapped n class="Chemical">1,3-dimethylcyclobutadiene is showclass="Chemical">n below iclass="Chemical">n Figure 5. It differs from the origiclass="Chemical">nal iclass="Chemical">n showiclass="Chemical">ng a thermally activated reactioclass="Chemical">n arrow coclass="Chemical">nclass="Chemical">necticlass="Chemical">ng the class="Chemical">n class="Chemical">1,3-dimethylcyclobutadiene 4 to 2. This last possibility is not explicitly discussed in the original report [25], although there is there an implicit assumption that this process is slow at the temperature of the experiment, 175K. The original article therefore seeks to persuade the reader on the basis of crystallographic evidence that the structure of 3 or 4 has been established, with the aid of the four colour figures included in the article, and (optionally for the reader) with acquisition of the CIF files.

The theme of the present article is to ask how a reader's experience and perception of a scientific article might be enhanced or simply altered by adopting new forms of presentation. In Figure 5, the relevant n class="Chemical">CIF file caclass="Chemical">n be loaded as a secoclass="Chemical">nd layer iclass="Chemical">nto the reactioclass="Chemical">n diagram. Because of the relatively large class="Chemical">number of host atoms iclass="Chemical">nvolved, the effect caclass="Chemical">n be somewhat overwhelmiclass="Chemical">ng wheclass="Chemical">n this is doclass="Chemical">ne aclass="Chemical">nd the iclass="Chemical">nterpretatioclass="Chemical">n may also be made more complex by the preseclass="Chemical">nce of disorder iclass="Chemical">n the guest. A further layer of iclass="Chemical">nterpretatioclass="Chemical">n caclass="Chemical">n be added by aclass="Chemical">nclass="Chemical">notaticlass="Chemical">ng the diagram with selected atom-atom distaclass="Chemical">nces; the reader caclass="Chemical">n use the display software to added further such aclass="Chemical">nclass="Chemical">notatioclass="Chemical">ns of their owclass="Chemical">n if they wish. There are maclass="Chemical">ny other actioclass="Chemical">ns the reader caclass="Chemical">n perform at this poiclass="Chemical">nt [19]. A further, this time smaller, alterclass="Chemical">native layer that coclass="Chemical">ntaiclass="Chemical">ns oclass="Chemical">nly the kerclass="Chemical">nel of the scieclass="Chemical">ntific problem (as perceived by the preseclass="Chemical">nt author, which may or may class="Chemical">not correspoclass="Chemical">nd to the perceptioclass="Chemical">n of the origiclass="Chemical">nal [25] authors) has beeclass="Chemical">n added here, aclass="Chemical">nd agaiclass="Chemical">n four key measuremeclass="Chemical">nt aclass="Chemical">nclass="Chemical">notatioclass="Chemical">ns made, togclass="Chemical">n class="Chemical">ether with selected bonds highlighted in a different colour.

The scientific problem can now be stated in the form of the following questions.

1. What are the kinetics of the reverse reaction of 4 to give 2 at 175K?

2. Does the crystallographic evidence convince that the guest is best described as n class="Chemical">1,3-dimethylcyclobutadiene iclass="Chemical">n close proximity to a detached molecule of class="Chemical">n class="Chemical">carbon dioxide?

3. More spen class="Chemical">cifically, how should the iclass="Chemical">nteractioclass="Chemical">n betweeclass="Chemical">n the labelled class="Chemical">n class="Disease">atoms C2 and C3 be interpreted? Should it be considered a strong van der Waals contact, as suggested by the original authors [25] or as a covalent bond? The same question might apply to another atom pair, O1 and C6 also connecting carbon dioxide and the cyclobutadiene.

4. Likewise, how should the angles n class="Disease">O1-C2-O7 or C2-C3-H10 be iclass="Chemical">nterpreted?

The reader may note a common theme emerging between these questions and the origins of helical stability in DNA as discussed above.

The first of these questions was in fact posed in the form of a blog, written by the present author [26] and based on chemical precedent and entropic arguments. It was posted in August 2010, little more than a month after the original report was first published. The precedent for this form of discourse when addressing a scientific issue had already been established [27,28]. Questions 2-4 emerged more conventionally and a little later in November 2010 in the same journal as the original article, and took the form of comments submitted by two independent groups [29,30]. The original authors have a right of reply to such comments, which they took [31]. These various n class="Species">participants iclass="Chemical">n the debate all had access to the same class="Chemical">n class="Chemical">CIF data as is transcluded into Figure 5. The debate to this point was summarized in a second blog post [32], and this and the original post themselves attracted ~15 responses in the form of appended comments. These posed further questions, on themes such as the computed structure of 1,3-dimethylcyclobutadiene, a debate on how much energy was required for angular distortion of O1=C2=O7 as an isolated molecule, and whether molecule 2 is transparent to light in the 320 to 500nm excitation range employed by the original experiments. This latter point was followed up by calculations of the UV-visible absorption spectrum of 2 inside the host cavity, also appended to the blog, and finally by calculations of the predicted vibrational spectra. The next stage in the discourse occurred in a conventional journal [12], taking the form of a set of calculations on the likely barrier preventing 4 and carbon dioxide from recombining inside the host cavity, and addressing question 1 above in more complete detail. This article did have one less conventional aspect; in the "rich HTML" version, an interactive version of the table of data was made available [33] in very much the manner adopted for Figures 3 and 5 in the present article. Additionally, there were links in this table to digital repository entries [34], which would enable any interested reader to access the complete archived details of all the calculations reported in that article.

Shortly after this last article was published, in December 2010, several publishers chose to highlight this emerging debate with editorial blog posts of their own [35-38]. These posts in turn attracted further comments, including several by one of the original authors. One comment in particular [39] entitled "Request of calculated structure data" highlighted an important aspect concerning the accessibility of previously reported data [12]. This alludes to the "rich HTML" table, and the observation that it is important to provide information on the file formats in which data is held, so that appropriate conversions if needed and concomitant visualization can be performed. This particular query was answered in the form of another blog post [19], and applies directly to the issue of how to re-use data associated with the current article (Figures 3 and 5).

The scientific discourse described above regarding the nature of the species in a host crystal lattice is still ongoing, and so a final consensus (if ever achieved) cannot be reported at the time of writing. It is noteworthy that the primary (hkl) crystallographic data relating to the original measurements has been provided upon request [40] and so further analysis of alternative crystallographic refinement models is now possible.

Conclusions

The two scientific examples discussed in this article span 57 years, a relatively short period in the history of the scientific journal. The first is arguably the most influential scientific article of the 20th century, and clearly the absence of data associated with it has not held back its recognition as such. What is also clear is that addition of such data, albeit 57 years after the original report, may have the potential to reveal further insights into the structure of DNA that may not have hitherto been highlighted. Whn class="Chemical">ether such a data-rich reformulatioclass="Chemical">n of the origiclass="Chemical">nal problem has aclass="Chemical">ny measure of impact remaiclass="Chemical">ns to be established. The secoclass="Chemical">nd article is oclass="Chemical">nly moclass="Chemical">nths old, but iclass="Chemical">n that brief period has beeclass="Chemical">n subjected to the kiclass="Chemical">nd of scruticlass="Chemical">ny that caclass="Chemical">n oclass="Chemical">nly be achieved by haviclass="Chemical">ng access to rich data sets. Oclass="Chemical">ne might fairly coclass="Chemical">nclude that the scieclass="Chemical">ntific article has evolved to eclass="Chemical">nable that scruticlass="Chemical">ny. The article that you are class="Chemical">now readiclass="Chemical">ng I suggest is oclass="Chemical">ne model for how such scieclass="Chemical">ntific discourse caclass="Chemical">n be both improved aclass="Chemical">nd accelerated. It remaiclass="Chemical">ns to be seeclass="Chemical">n if scieclass="Chemical">ntists are prepared to author such articles iclass="Chemical">n the future. There is aclass="Chemical">n early example [41] of aclass="Chemical">n article where both the discourse aclass="Chemical">nd the data supporticlass="Chemical">ng that discussioclass="Chemical">n were seamlessly iclass="Chemical">ntegrated iclass="Chemical">nto oclass="Chemical">ne (XML-based) documeclass="Chemical">nt, with the preseclass="Chemical">ntatioclass="Chemical">n beiclass="Chemical">ng made available to the reader by applicatioclass="Chemical">n of suitable stylesheet-based traclass="Chemical">nsformatioclass="Chemical">ns. The productioclass="Chemical">n of aclass="Chemical">n article iclass="Chemical">n this form was however class="Chemical">noclass="Chemical">n trivial. Siclass="Chemical">nce theclass="Chemical">n tools have appeared to facilitate the process [42,43] aclass="Chemical">nd the task class="Chemical">now much be to reach both the hearts aclass="Chemical">nd the miclass="Chemical">nds of scieclass="Chemical">ntific authors to eclass="Chemical">ncourage them to start adopticlass="Chemical">ng this form of eclass="Chemical">nhaclass="Chemical">nced scieclass="Chemical">ntific discourse.

Competing interests

The author declares that they have no competing interests.

Additional file 1

Interactive Jmol-enhanced version of Figure 3.

Click here for file

Additional file 2

Enhanced version of Figure 4containing additional hyper links (This figure should be viewed with a web browser capable of SVG display, such as Chrome, FireFox, Safari or IE 9).

Please note: The figure is not currently displayed as intended by the author due to technical issues with the BMC site. This will be resolved as soon as possible.

Click here for file

Additional file 3

Interactive Jmol-enhanced version of Figure 5.

Click here for file