An attempt to map the NLM Journal Publishing DTD to XHTML while preserving as much information as possible.
Notes:
- Lots of minor elements are left out, for clarity.
- The citation microformat is still being discussed.
- Authors and addresses should all be hCard, eventually.
- This is what (X)HTML was designed for, after all.
<div id="article">
<div id="article-front">
<h2 id="article-title">ARTICLE TITLE</h2>
<ul class="authors">
<li class="author">AUTHOR<a class="xref" href="#aff1">1</a></li>
<li class="author">AUTHOR<a class="xref" href="#aff2">2</a></li>
</ul>
<ul class="affiliations">
<li class="affiliation" id="aff1">ADDRESS</li>
<li class="affiliation" id="aff2">ADDRESS</li>
</ul>
<div id="abstract">ABSTRACT</div>
<div class="citation self">CITATION</div>
</div>
<div id="article-body">
<div class="section" id="introduction">
<h3>Introduction</h3>
<p id="para-1">PARAGRAPH</p>
<p id="para-2">PARAGRAPH[<cite><a href="#ref1" class="xref bibr">1</a></cite>]</p>
<div class="section" id="section-1.1">
<h4>SECTION TITLE</h4>
<p id="para-3">PARAGRAPH</p>
<p id="para-4">PARAGRAPH</p>
</div>
</div>
<div class="section" id="methods">
<h3>Methods</h3>
<p id="para-5">PARAGRAPH</p>
<p id="para-6">PARAGRAPH</p>
</div>
<div class="section" id="results">
<h3>Results</h3>
<p id="para-7">PARAGRAPH <a href="#table1" class="xref table">Table 1</a></p>
<div class="table-wrap" id="table1">
<table class="data">
<caption>
<span class="item-number">Table 1.</span> CAPTION</caption>
<col/>
<col/>
<thead>
<th>
<td>HEADING</td>
<td>HEADING</td>
</th>
</thead>
<tbody>
<tr>
<td>CELL</td>
<td>CELL</td>
</tr>
<tr>
<td>CELL</td>
<td>CELL <span class="note"><a href="#table1fn1" class="xref table-fn">*</a></span>
</td>
</tr>
</tbody>
<tfoot/>
</table>
<div class="footnotes">
<p class="footnote" id="table1fn1"><sup>*</sup> FOOTNOTE</p>
</div>
</div>
<p id="para-8">PARAGRAPH <a href="#figure1" class="xref fig">Figure 1</a></p>
<div class="figure" id="figure1">
<a rel="enclosure" href="IMAGE-URL"><img class="figure" src="IMAGE-URL" alt="Figure 1"/></a>
<div class="caption"><span class="item-number">Figure 1.</span> CAPTION</div>
</div>
</div>
<div class="section" id="discussion">
<h3>Discussion</h3>
<p id="para-1">PARAGRAPH</p>
<p id="para-2">PARAGRAPH</p>
</div>
</div>
<div id="article-back">
<div id="acknowledgements">
<h3>Acknowledgments</h3>
<p>PARAGRAPH</p>
</div>
<div id="appendices">
<h3>Appendices</h3>
<div class="appendix">
<h4>APPENDIX-TITLE</h4>
<a rel="enclosure" href="SUPP-DATA-URL">SUPP-DATA-TITLE</a>
</div>
</div>
<div id="abbreviations">
<h3>Abbreviations</h3>
<dl>
<dt>TERM</dt>
<dd>DEFINITION</dd>
<dt>TERM</dt>
<dd>DEFINITION</dd>
</dl>
</div>
<div id="bibliography">
<h3>References</h3>
<ol id="reference-list">
<li class="citation article" id="ref1">
<span class="title"><a class="url" href="URL">ARTICLE-TITLE</a></span>.
<span class="author vcard n">AUTHOR</span>.
<span class="container"><span class="title">JOURNAL-TITLE</span> <span class="year">YEAR</span>; <span class="volume">VOLUME</span>(<span class="issue">ISSUE</span>):<span class="page-start">PAGE</span></span>.
Accessed <abbr class="date-accessed" title="YYYY-MM-DD">DATE</abbr>
</li>
</ol>
</div>
</div>
</div>
Comments
All fields are optional, email address will not be shown; no HTML, URLs are automatically hyperlinked.

Looks good. The fun part is going to be getting authors to use it, you realize. :)
This will also be a good template for publishing articles (for use with microformat and GRDDL processors), so hopefully it'll be useful even if it's hard to get people to author in this format initially (which I'm not sure it will be necessarily, as long as there are some good templates).
won't it make sense to start an "article" microformats in microformats.org? since they already had an book proposal.
Yes, I'll try and do that at some point - I already mentioned it on the list once. The trouble is that microformats prefer to use class names, whereas for this purpose it makes more sense to use unique ids (as there's only ever going to be one article on a page at a time).