chem-bla-ics

Minting RDF from CSV files with Bioclipse #MIILS2013

2013-06-18T16:31:00.000+01:00

The below slides are part of the introduction for the hands on for the #MIILS2013 course this afternoon. I had the participants look at creating RDF the hard way: using Bioclipse script (using this Bioclipse-OpenTox version). And, they had to follow the Open PHACTS RDF Guidelines, VoID specification, etc.

#MIILS2013 Practical: Bioclipse for RDF minting from Egon Willighagen

The Death of the Single Pass Peer Review

2013-06-14T09:50:00.002+01:00

One key aspect of peer review is to ensure that the paper is scientifically sound. The publishing system therefore asks peers to review the paper and test that. Part of this is to see if relevant literature is cited, backing up key finding around the addressed hypothesis in the paper. After all, the research must be novel.

However, there is so much literature nowadays in so many different journals, a phenomenon initiated by the large publishing industries, that it has become hard to track things. No worries, these same industries have come up with tools to handle that. A second problem is that research has become that interdisciplinary that even focusing on a few journals of which you can even read the papers easily is hard.

Consequently, it is easy to miss even key papers. He who reads all relevant, important literature cast the first tweet. Thus, the peer reviewers jump in, and make you aware if you missed something critical.

Now, these peers actually have the same issue. Worse, because it is hard to find good reviewers, we settle with post-docs and peers not specialist in the topic of the paper. Of course, we'll find plenty of reasons why this is good. But the bottom line is, that even peer review fails to address all critical points in the manuscript. And, as a result, we can all point to literature in the glossies where important aspects have been neglected (he who never found a "high impact" paper without flaws, cast the first researchblogging.com post), and an increase in the number of retractions recently.

Therefore, what the publishing community needs is to admit that the current approach is no longer sufficient. It worked well, for some 40-50 years, we did not need it before then, and it no longer scales with the output. I am not implying that peer review is bad, but the current single pass peer review implementation is.

Instead, I call on all publishers to step away from the current implementation, and adopt a multistep peer review process. Possible approaches include:

only accept papers that have appeared in a pre-print server
implement a two-pass system, with a quick first review scanning if key literature has been discussed, just covering the introduction
open peer review

I know that some of these solutions are being experimented with. All I ask is for publishers to strongly support this, and demand from editors and authors to do the same.

Why? Because your journal quality will actually improve.

Can I haz thiz PDF, pleaz? #icanhazpdf

2013-05-20T10:26:00.000+01:00

Readers of my blog know that I am a promoter of Open Science and have been so for a very long time. One of the main reasons why I am, is that Open just is more efficient. No worry about getting formal approval to access something, faster exchange of knowledge, and people just doing their science.

Situation #1
You published your paper and it is appealing to many people. They all start emailing you for a reprint and you end up spending an hour or two every week answering those emails and attaching the PDF to your replies.

Situation #2
Not that you really have time for it, but don't you just wish you had time to print that really interesting paper to put it on your desk pile of papers to read? Well, no luck, your library had to cancel the subscription. And IBL is just a too expensive approach (and sometimes slow).

Now, because the majority of my papers are Open Access, I never have Situation #1. In fact, I don't think anyone gets such emails a lot. (Please leave a comment if you get more than one such request in half a year.) Situation #2 is much more common, though with access to two university libraries, the journals must be really obscure for me not to have access. But there are additional options to solve Situation #2 before you make an author run into Situation #1.

The upcoming solution to this (well, one of them, there are others) is using #icanhazpdf. Just use this hashtag on social media to let people know you really like to read that paper. The screenshot of a Twitter search shows some examples. This is definitely a faster mechanism to have access to reprint papers.

Now, is this legal? No (unless one of the original authors replies by sending a reprint). Do I recommend it? No. See e.g. this post. Is #icanhazpdf part of Open Science? No. It is a symptom of the limitations in scalability of closed-access publishing. With so many people still dying of things that we are so close to curing, I believe scalability is of critical importance.

Therefore, I rather solve this literature access issue by publishing in (true) Open Access journals. You should too; it really is not that expensive (e.g. comparable to one ACS visit and often cheaper). But #icanhazpdf is an interesting trend, one that I feel you should be aware of. And I blog this, because not everyone seems to know this alternative yet.

BTW, another excellent alternative, particularly in The Netherlands. Visit a friend at another university in the afternoon/evening and spend the rest of the day in the library of the other university to copy (download) those interesting papers. The IBL systems will tell you exactly which journals the other university has. Slower, but comes with the advantage of giving you an excuse to visit your friends elsewhere.

Angelina Jolie's breasts, SNPs, and Jmol

2013-05-16T23:08:00.000+01:00

I never had thought I would ever blog about Angelina Jolie's breasts, but they were the news this week that turns out to help me answer a student question for our bioinformatics course. The question was about how to visualize SNPs in 3D protein structures. The link is the protein encoded by BRCA1. Some SNPs in BRCA1 are strongly (cor)related to cancer, as you can read about in OMIM. One 3D protein structure in the pdb.org is 1JM7, which is a solution based on NMR experiments by Brzovic et al. (doi:10.1038/nsb1001-833).

The Sequence tab for this structure provides us the sequences of the two proteins in this structure, one of which is BRCA1. We can add the SNP annotation with the drop down box:

(BTW, not all structures have SNP information, so this drop down box option is not always present!)

It will show that very many SNPs are known. OMIM needs to be checked to see if they are associated with disease. It should look like this:

You can see the "Single Nucleotide row, with many colored SNPS circles. The first is Met1 and the second is Arg7. This post is, however, on how to visualize the SNPs in Jmol. Now, the Jmol applet is linked to on the pdb.org website, on the right side in the same Sequence tab:

If you click the "Display Jmol" option, the Jmol applet should fire up (your browser does require the Java plugin). That will create a small dialog hovering over that corner with the Jmol applet. You can right-click on the background and select the Console option:

That will fire up another dialog window, one that will look like the one below and in the lower large section we can write command (yes, some minimal programmaing, in Jmol script!). For example, we can select the methionine at position 1 (the first SNP from BRCA1 as found earlier):

Where select Met1 is typed, you type that too, and press enter. You have now selected 266 atoms as will be written in the top large text area. Then you type the following two further commands, and notice the effect in the Jmol applet window, creating a set of three commands:

select Met1
spacefill on
color green

Then, for each SNP you can pick a different color, and easily visualize a set of SNPs. Now, among all the variants it reports missense SNPs for Cys61 and Cys64. So, the following six commands will visualize these two SNPs, with Cys61 in red, and Cys64 in green:

select Cys61
spacefill on
color red
select Cys64
spacefill on
color green

There are some additonal zinc ions in the structure, which you can color orange with:

select zinc
color orange

This will create this view in Jmol:

So, you can clearly see that the two cysteines affected by the SNPs are involved in the structure's zinc finger. No wonder those mutations give trouble.

Have fun!

Brzovic, P. S., Rajagopal, P., Hoyt, D. W., King, M.-C. & Klevit, R. E. Structure of a BRCA1–BARD1 heterodimeric RING–RING complex. Nature Structural & Molecular Biology 8, 833-837 (2001). URL http://dx.doi.org/10.1038/nsb1001-833.

Book Review: "Instant Cytoscape Complex Network Analysis How-to"

2013-05-12T09:59:00.002+01:00

Instant Cytoscape Complex Network Analysis How-to (ISBN:978-1-84951-980-9) is a book by Gang Su and reviewed by John Morris, which currently costs less than 10 euro. It is a thin book, with a mere 76 pages of which 58 pages are really informational. Of those pages, a lot of space is taken up by screenshots, which you may or may not like.

In the ten chapters of the book, it does give a concise introduction to Cytoscape. It is clearly written and easy to digest. The chapter titles read as follows, giving you an idea of the topics it covers:

Unboxing Cytoscape (easy), Loading up Cytoscape! (easy), Touch Cytoscape with some style, A network with many forms, Finding needles in a haystack, Tuning up Cytoscape with gadgets, Where are the clusters? (advanced), Additional visualization capabilities, When local is not enough, Export, save, and call it a day!

The qualification of easy and advanced looks somewhat arbitrary. I would guess for many scholars, installing the software may in fact be more difficult than pushing some buttons to do a cluster analysis. Otherwise, the book is targeted at people whit basic programming knowledge, network analysis, and Cytoscape usage. The requirements do not list that MS-Windows is needed, but assumes that is the case. That puts OS/X and Linux users at a disadvantage, particularly with the installation chapter.

The [PACKT] publishers provides the book in various formats, and I mostly read the PDF version. I did have a quick look at the EPUB version. Compared to PDF this format has the advantage the text shown will scale to fit the device, whereas a PDF page is always the same. Unfortunately, this version lost some layout aspects, leading to confusing text, badly shown URLs, and only partly shown screenshots in the Aldiko Book Reader. But because whitespace is one of the important things, I would recommend the PDF. However, you can download it in multiple formats, making this a smaller issue.

Each chapter consists of one or more sections called “Getting started”, “How to do it...”, “How it works...”, “There’s more...”, “See also”. Practically, however, you run through a chapter on one go, perhaps also, because of the (limited) length of chapters. I would very much recommend more of the "There's more..." extended from small hints to full texts, giving the book both more mass and the author more justice.

In fact, compared to tutorials available online, the book mostly provides the convenience of having the information in a nice narrative. However, things that a book should provide too is an index, but that is missing from the book. I know I can use ^F too, but that is not entirely the same thing. An index has more power in guiding the user the the right section of the book. That is, when I browse an index, I see other words the author found important. It is like a table of content, and chapter titles you can also find with ^F after all.

The examples are easy to follow, and assuming Cytoscape 2.8 does not change a lot within one series, we can assume the screenshots and instructions will not change. That is always the risk with a book like this. I hope the book will see frequent extensions and updates.

There were a few things that I had trouble with. One thing is the lack of distinction between Open Source and free of cost. Those are not the same thing. So, rather than writing "It is open source (free of cost)..." I would very much prefer it to read "It is open source and free of cost...". Also, a categorization of MS-Excel as a text editor is awkward, and I hope the next edition will also use OpenOffice/LibreOffice as replacement which is way more in line with the "It is open source..." introduction in the first place.

Content wise, I think the book is fairly solid and it surely covers all the basics one may want to do. A few plugins are mentioned, though they are a bit hidden. I would very much like to see a few plugins appear as separate (advanced?) chapters as show cases for how all that basic functionality is put to use. This is in parallel with the "there is more" sections, which could refer to such chapters. A bit of work, but it would make the book a lot more valuable and go beyond the basic tutorial nature it has now.

All in all, I think the book has a very reasonable price/content ratio and combined with the fact that some environments live by if-it-is-not-in-the-library-it-does-not-exist, I can recommend to keep an eye out for this book. Perhaps wait for the next edition, adding at least an index, but it is an easy and cheap way to support and encourage the Cytoscape community.

Oh, and while composing the below CiteULike widget, I realized that the book lacks references to other literature. For example, there is no reference to the Cytoscape articles. That would be another way to add value over the online tutorials.

Su, G. Instant Cytoscape Complex Network Analysis How-to (2013). ISBN:978-1-84951-980-9.

Groovy Cheminformatics with the CDK - 8th edition

2013-05-11T21:35:00.003+01:00

Update: I have uploaded the full ToC to FigShare.

It is still not what I want to release on Amazon.com as a final version, but fewer and fewer bits are, what I consider, missing. And with a whopping 228 pages I am happy I have a solid build system where I can just hack in new code and update the CDK version easily. In fact, this version adds 10 new scripts since the 7th edition. Some scripts take more time than others, and four of these are solutions I wrote for the Chemistry Toolkit Rosetta (CTR) by Andrew Dalke (see also this blog post).

There are two versions:

paperback, for $ 45
eBook, for $35, a PDF version

Still, this revision does not add that much new content:

Section 2.2.2. Bond stereochemistry (just refers to Chapter 3)
Chapter 3. Stereochemistry
Chapter 19. Chemitry Toolkit Rosetta
Section 21.3.2. Grabbing dependencies (for Groovy)

The start of the stereochemistry chapter is interesting, explaining wedge bonds and the ITetrahedralChirality interface, but obviously I have to add the IDoubleBondStereochemistry will have to follow soon.

The first words of Chapter 19 look like what I tumblred earlier today (yes, I have an introductory chapter upcoming but not ready yet, causing the chapter numbering to differ):

The CTR chapter has solutions of four of the challenges listed in the wiki. I have not yet managed to create solutions for all 18 of them, as intended.

I still like to stress that the book is for convenience: it only groups together information that you can find elsewhere as well. For example, the solutions to the four CTR problems can be found in that wiki too. The solution is there, and the explanation is in the book.

CDK 1.4.18: the changes, the authors, and the reviewers

2013-05-11T12:53:00.001+01:00

The development of CDK 1.4 is really slowing down, and the CDK 1.5.2 release by John shows how far the development in the master branch has picked up. (BTW, you do know Planet CDK, right? The perfect way to keep up with CDK development!)

This patch does not have a killer fix, but some things of interest. One patch updated the InChI to structure algorithm to also set atomic numbers for atoms, a patch that ensures the "C" and "C" are isomorphic (one of this use cases for which the algorithm was never written to work), hydrogen isotope reading from MDL molfiles, and passing of uncertain double bond stereochemistry to the InChI generator.

In all cases, users of CDK 1.4.x versions are recommended to upgrade to this latest stable version.

The changes

Added a missing dependency 6ee0f0a
setting unspecified bonds when generating and InChI 631281a
unit test for bug1295 724366e
Bumped the copyright year to 2013 22ff5a9
Unit test to verify that wedge bond information is properly read ea8a316
Unit tests for the expected behavior of bug #1294 c3a6452
Removed redundant import (not found in class path) bfb2b1b
Removed unused imports bc6e1a6
Added citation for permutation method 1edc2d5
Cleaned up documentation, reworded and added example usage. Update tutorial link to a site which has backed up the now missing web page. d76cac6
Fixed proton isotope perception. 421345b
replaced cast to implementation with a cast to the interface c273930
Small patch for bug 3551478 71f40a9
Added additional test that non matching symbols mismatch c532191
non-null atomic numbers when reading from InChi [bug:1293] 3117f4c
Test to ensure that the atom type name is returned with IAtomType.toString() 471a20c

The authors

As you can see below, this release has a patch by a new contributor, Magda Oprian.

8 Egon Willighagen
8 John May
1 Magda Oprian
1 Stephan Beisken

The reviewers

7 Egon Willighagen
4 John May

Linking WikiPathways to binding affinity data

2013-05-10T13:31:00.001+01:00

WikiPathways is a long established project (doi:10.1371/journal.pbio.0060184), and while in beta provides a rich resource on various biological processes, in various species. Open PHACTS (Andra particularly) is working on integrating this data. Obvious links include the proteins listed in pathways, which are targeted by compounds in, for example, ChEMBL, one of the resources in the Open PHACTS cache.

However, drugs and drug-like molecules are not abundant in WikiPathways. Some are, like morphine in this metabolism pathway:

However, it seem more common that drug-like compounds are indirectly found pathways. For example, angiotensin-converting enzyme inhibitors (ACE inhibitor, in red below) are mentioned in pathways:

An example ACE inhibitor is captopril. The current BridgeDB-based identifier mapping (doi:10.1186/1471-2105-11-5) will not link the ChEBI identifier for ACE inhibitors (ChEBI:3380) to the matching ChEMBL entry (CHEMBL1560). Thus, we cannot make this link directly.

However, we're in a semantic world, and ontologies come to the rescue, in particular, the ChEBI ontology (doi:10.1093/nar/gkm791). The following screenshot shows how ChEBI says captopril is an ACE inhibitor:

So, if we can only make a claim that any of those listed specific drugs is a ACE inhibitor, then we can use that information to link captopril to the WP557 pathway.

Christian Brenninkmeijer and Alasdair Grey have introduce the concept of lenses in Open PHACTS (see paper below) that define when two things are considered the same. For example, under certain conditions protonation states are the same, while on others not. By turning on and off lenses, you can define what is the same. Lenses can be grouped, and we could have, for example, a "human cell" lens the groups all lenses that details what chemicals are "biologically" identical. For example, because the are readily interconverted under biological (cell) conditions.

So, ChEBI defines a (one-directional) lens that defines captopril is the same as "ACE inhibitor".

Pico, A. R. et al. WikiPathways: Pathway editing for the people. PLoS Biol 6, e184+ (2008). URL http://dx.doi.org/10.1371/journal.pbio.0060184.

van Iersel, M. et al. The BridgeDb framework: standardized access to gene, protein and metabolite identifier mapping services. BMC Bioinformatics 11, 5+ (2010). URL http://dx.doi.org/10.1186/1471-2105-11-5.

Degtyarenko, K. et al. ChEBI: a database and ontology for chemical entities of biological interest. Nucleic acids research 36, D344-D350 (2008). URL http://dx.doi.org/10.1093/nar/gkm791.

Brenninkmeijer, C. et al. Scientific lenses over linked data: An approach to support task specific views of the data. a vision. In Linked Science 2012 - Tackling Big Data (2012). URL http://ceur-ws.org/Vol-951/paper5.pdf.

New Paper: "The ChEMBL database as linked open data"

2013-05-09T13:45:00.002+01:00

Update: Mark wrote up a blog post on the RDF that the ChEMBL team itself.

Yesterday, the paper "The ChEMBL database as linked open data" (doi:10.1186/1758-2946-5-23) by Andra Waagmeester (@andrawaag), Ola Spjuth (@ola_spjuth), Peter Ansell (@p_ansell), Antony Williams (@chemconnector), Valery Tkachenko, Janna Hastings, Bin Chen (@binchenindiana), David J Wild (@davidjohnwild), and me appeared in the OA JChemInf journal.

I am also indebted to the ChEMBL team (@chembl) for both providing such valuable data under a liberal Open Access license and their critical reading of the manuscript! Additionally, I would like to stress that the ChEMBL team will create their own RDF version of ChEMBL and that this paper is not describing the version they will release.

BTW, the source of the paper is available from GitHub. And the (original) scripts to create RDF from the MySQL dump of ChEMBL are also on GitHub.

This paper outlines the RDF as it has evolved from various earlier projects. The above diagram visualizes the basic structure (red), various Linked Data resources linked too (blue) and illustrates how various ontologies are used, such as the CHEMINF, BIBO, and CiTO ontologies.

Additionally, various applications and links are described developed by various co-authors. For example, Peter worked on the use in Bio2RDF and Bin and David on Chem2Bio2RDF. Andra developed an extension for his (#altmetric) CitedIn resource, giving credit to a paper when data in it is extracted into ChEMBL. Ola, Valery, and Anthony developed a Bioclipse Decision Support extension, which supports a nearest neighbor search in ChEMBL using ChemSpider. Of course, Ola also hosts the SPARQL end point of which you can monitor the uptime at the also cool mondeca.com service:

(Yes, I think I have all the cool buzzwords covered in this paper. Sadly, marketing is needed nowadays as a scientist. Where is the time that you could rant on page after page in all your domain specific jargon, not having to worry if your reader would understand it immediately, or without a university degree...)

What this paper does not describe, is all the things I did with ChEMBL-RDF in the Open PHACTS project (@Open_PHACTS), which includes the use of QUDT and the jQUDT library for unit normalization outlined in this document and the use of VoID for link sets as described in this document.

Willighagen, E.; Waagmeester, A.; Spjuth, O.; Ansell, P.; Williams, A.; Tkachenko, V.; Hastings, J.; Chen, B.; Wild, D. Journal of Cheminformatics 2013, 5, 23+.

BridgeDB identity mapping file for metabolites

2013-05-06T10:09:00.000+01:00

Based on HMDB 3.0 (January data; see the reference at the end), I have updated the identity mapping file for metabolites for BridgeDB. It can be downloaded here as pre-release. I here document how I created this file. I have placed the script online at GitHub so that anyone can look at the code and make custom BridgeDB Derby files too. (For example, for large CAS registry number mapping files.)

The code is some 100 lines long. The input .zip is processed by iterating over the containing XML files:

zipFile.entries().each { entry ->
if (!entry.isDirectory()) {
println entry.name

inputStream = zipFile.getInputStream(entry)
def rootNode = new XmlSlurper().parse(inputStream)

String rootid = rootNode.accession.toString()
}
}

For each file it creates an entry in the BridgeDB data:

Xref ref = new Xref(rootid, BioDataSource.HMDB);
error = database.addGene(ref);

// add the synonyms
addAttribute(
database, ref, "Symbol",
rootNode.common_name.toString()
);
addAttribute(
database, ref, "Synonym",
rootNode.traditional_iupac.toString()
);
addAttribute(
database, ref, "Synonym",
rootNode.iupac_name.toString()
);

// add the SMILES, InChIKey, etc
addAttribute(
database, ref, "InChIKey",
cleanKey(rootNode.inchkey.toString())
);
addAttribute(
database, ref, "SMILES",
rootNode.smiles.toString()
);
addAttribute(
database, ref, "BrutoFormula",
rootNode.chemical_formula.toString()
);
addAttribute(
database, ref, "Taxonomy Parent",
rootNode.direct_parent.toString()
);

These are some basic properties, and adding the mappings looks like:

// add external identifiers
addXRef(
database, ref,
rootNode.accession.toString(), BioDataSource.HMDB
);
addXRef(
database, ref,
rootNode.cas_registry_number.toString(), casDS
);
addXRef(
database, ref,
rootNode.inchi.toString(), inchiDS
);
addXRef(
database, ref,
rootNode.chemspider_id.toString(), chemspiderDS
);
addXRef(
database, ref,
rootNode.pubchem_compound_id.toString(), pubchemDS
);
addXRef(
database, ref,
rootNode.chebi_id.toString(), chebiDS
);
addXRef(
database, ref,
rootNode.kegg_id.toString(), keggDS
);
addXRef(
database, ref,
rootNode.wikipedia.toString(), wikipediaDS
);
addXRef(
database, ref,
rootNode.drugbank_id.toString(), drugbankDS
);
addXRef(
database, ref,
rootNode.nugowiki.toString(), nugoDS
);

You can thus see that this new metabolites identity mapping file has mappings to various data sources, including ChemSpider, PubChem, ChEBI, KEGG, Wikipedia, DrugBank, and a few others. All these mappings are provided by HMDB 3.0.

The Derby database file is the format that PathVisio can read natively. This is the magic that is used in the script to create and finalize (aka save) this file:

GdbConstruct database = GdbConstructImpl3.createInstance(
"hmdb_metabolites", new DataDerby(),
DBConnector.PROP_RECREATE
);

database.createGdbTables();
database.preInsert();

String dateStr = new SimpleDateFormat("yyyyMMdd").
format(new Date()
);
database.setInfo("BUILDDATE", dateStr);
database.setInfo("DATASOURCENAME", "HMDB3");
database.setInfo(
"DATASOURCEVERSION",
"hmdb_metabolites_" + dateStr
);
database.setInfo("DATATYPE", "Metabolite");
database.setInfo("SERIES", "standard_metabolite");

// process the zip file

database.commit();
database.finalize();

There are a few more details, but nothing not exposed by the full script.

Wishart, D. S. et al. Nucleic Acids Research 2013, 41, D801-D807.

OpenTox Euro 2013: call for abstracts

2013-04-27T11:58:00.000+01:00

Later this year OpenTox Euro 2013 will take place in Mainz:

I also created a page on Lanyrd where you can enlist yourself, preferably with your Twitter account. BTW, the hashtag to be used is #oteu13:

You can submit abstracts here latest by 30 June 2013. There will be sessions around these optics: Open Infrastructure and Application Development, Integrated Data Analysis, Visualisation, Cheminformatics, Bioinformatics, and Systems Biology.

Must-have-ORCID reason #314

2013-04-27T09:43:00.001+01:00

Publisher writes:

2. Author details: Author "...@maastrichtuniversity.nl" provided in the manuscript differs from the one reflected in the submission system which is "..@gmail.com". We have proceeded and followed the manuscript. Please check and advise if action taken is appropriate.

Yes, that is both me.

#ACSNola talk: "An architecture for an Open Science molecular compound database"

2013-04-20T12:25:00.001+01:00

About half a year ago I was fed up with the slow progress in Open Data in chemistry. Some initiatives exist and some projects, but there is no clear central point of access, which particularly is a problem for smaller providers. Thus, I set out a project plan to make this change. There were two other aspects that I wanted to include here:

licensing must be explicit, to allow aggregators to know under what conditions they can redistribute that data (or not)
compound databases must start being clear on whether entries are specific compounds and if listed properties are for a specific tautomer (or not)

This is of critical importance to do reasoning over data in multiple data sets, as recently outlined in our Applications of the InChI paper, or for large data integration projects like Open PHACTS.

This presentation captures all the usual suspects, like the Panton Principles, lists some truly Open Data in chemistry (e.g. CrystalEye), and outlines the architecture I am working on. The primary purpose of this project is Linked Open Data for chemistry and to boost this field. Sadly, grant writing interfered with my agenda, and I did not manage to complete the full demo, but the slides contain this real-world screenshot that shows what it looks like (and I expect this put this publicly online in 1-2 months):

By no means this architecture expected to be as functional as Open PHACTS or to replace large compound databases like ChemSpider or PubChem. Instead, it is meant as a simple architecture that does two things right and is simple enough to set up, that any chemistry lab can do it. Goal: to increase the size of the chemical Linked Open Data network, which is way too small at this moment. I will list LinkedChemistry.info data sets with DataHub.io.

Basically, you set up a SPARQL endpoint with the data you want to share and the Chemical Compound Box as PHP front end using ARC2. That's it.

And the slides of the #ACSNola presentation:

An architecture for an Open Science molecular compound database from Egon Willighagen

#ACSNola talk: "Bioclipse-OpenTox: interactive predictive toxicology"

2013-04-19T21:31:00.002+01:00

My third #ACSNola talk (well, second chronologically):

Bioclipse-OpenTox: interactive predictive toxicology from Egon Willighagen

John May is now release manager of CDK 1.5.x

2013-04-19T12:38:00.002+01:00

After years of being release manager, John May (from Chris' group) has started as release manager for the 'master' branch, and thus leading to CDK 1.5.x versions (and later 1.6.x).

As part of that, John now has commit powers, to push reviewed patches into the main source tree (like Rajarshi and I have too):

This role of gatekeeper is very practical and comes with a large responsibility: ensure that whatever it pushed actually compiles for everyone. The gatekeeper will get up in the middle of the night when the official repository's branches does not compile! No, but really, really close to that. It's has main priority, so we always double check with 'git clean dist-all test-dist-all' before we push. No, we actually now and then forget to do this, e.g. after porting patches from another branch. But we fix that immediately! Promised!

And, John pushed his first two patches to master (ported from cdk-1.4.x):

#ACSNola talk: "Open PHACTS: meaningful linking of preclinical drug discovery knowledge"

2013-04-18T11:11:00.002+01:00

Half a year ago I submitted this abstract for the #ACSNola meeting last week (and as in the slides, I stress that is a large community effort involving not only academic groups but also many pharma companies):

Open PHACTS: meaningful linking of preclinical
drug discovery knowledge

E. Willighagen, C. Brenninkmeijer, C. Evelo,
L. Harland, A. Gray, C. Goble, A. Waagmeester,
A. Williams

Recently, semantic web technologies have been
adopted by the life sciences community for this
purpose. However, while these new technologies
provide us with methods, they do not provide us
with an exact solution. Open PHACTS uses these
methods to solve problems in linking preclinical
knowledge from databases like Uniprot, ChEMBL,
and WikiPathways. Problems that are discusses
and for which our solutions will be presented
include: 1. approaches to map data between the
databases using the Vocabulary of Interlinked
Dataset, including identifier mapping with
BridgeDB, appropriate choices of mapping
predicates, and ontologies to cover provenance,
such as the Provenance Authoring and Versioning
ontology; 2. deal with different units for
experimental data using the Quantities, Units,
Dimensions and Data (QUDT) ontology for (on the
fly) quantity conversion; and 3. how all this
is linked to user-oriented graphical user
interfaces.

I have now uploaded the slides:

Also note the associate partnership program: it is not too late to join the 40 other associate partners and team up with Open PHACTS!

#ACSNola talk: "Correlating time spent on exercises with exam results in Protein Structure education"

2013-04-15T15:54:00.000+01:00

I have cleaned up the slides a bit, adding some information I explained during the talk in New Orleans.

Correlating time spent on exercises with exam results in Protein Structure education from Egon Willighagen

I'm a proud "RSC eScience hero"

2013-04-09T06:06:00.001+01:00

A while ago Antony Williams, of ChemSpider, was awarded an eScience price for his work in the field. He decided to put that money to use to award several people. He worked things out with the Royal Society of Chemistry who send out a press release a few days ago ("RSC eScience heroes rewarded through Microsoft prize"). And I am proud to be one of those.

A full list of winners is extracted from the press release:

Enormous thanks to Tony for this and congrats too to the other winners. I note that several of the winners are or have been involved in the Blue Obelisk movement, and seeing all the hard work rewarded by this prize indicates we are making impact. Personally, I am awarded for my work on the CDK particularly (but have worked on way more open source code for cheminformatics and statistics). I do stress that the CDK is a community project which I happen to be a long time contributor and founder of. But the success of the CDK is based on the effort from all authors and users!

If you are wondering about the impact of just the CDK, have a look at the growing list of papers citing the CDK papers, which I am tracking in CiteULike, which has good CiTO support (the detail):

Dr Evan Bolton to speak at Maastricht University: "PubChem: A platform for chemical biology"

2013-03-30T15:19:00.000+01:00

Mon 13 May Dr Evan Bolton of PubChem will visit our BiGCaT research group and will give a presentation on the PubChem small compound-bioassay database at 11:00am .

It is very much appreciated if you returned me an email confirming your attendance (firstname dot lastname @ maastrichtuniveristy dot nl).

PubChem: A platform for chemical biology

PubChem is an open archive for small molecules and their biological activities located at the National Center for Biotechnology Information (NCBI). Despite humble beginnings, PubChem continues to receive broad community support through a continued influx of new information and new information resource types. Over the past eight years, PubChem has seen dramatic growth in participation by the chemical biology community, in terms of contributors (averaging 25% year-over-year growth) and users (averaging 20% year-over-year growth). PubChem contents include more than 45 million small molecules (+1.8 million with biological testing results), 115 million substance descriptions, 640 thousand biological assay descriptions (against +5800 declared unique protein sequence targets), and over 200 million assay outcomes (a substance tested in an assay is an outcome) from more than 200 contributors. PubChem provides a number of tools to help navigate this vast corpus of information. In addition, PubChem integrates contents to other resource of chemical biology interest. As the needs of the community have changed, so too has PubChem adapted. This talk give an overview of the PubChem resource as it exists today with an emphasis on recently introduced features.

CDK 1.4.17: the changes, the authors, and the reviewers

2013-03-23T15:07:00.002+01:00

Hot on the heels of the CDK 1.4.16 changelog, here are the changes in CDK 1.4.17. The primary purpose of this release is to fix a regression in 1.4.16 for sorting IAtomContainers, which turned out a bit nasty.

Readers who are still using 1.4.16 or earlier are strongly encouraged to upgrade to this release.

The Changes

Identical copies of the same method in AbstractAtomContainerSet, but using IMolecule, solving an exception complaining about IMoleculeSet to contain only IMolecule [only for cdk-1.4.x] 67bed79
Added a unit test for the sorting of the multiplier ff2d820
updated sort to also sort multiplier values - bug:1291 d48d30d
renamed variable 'minimum' and added explanation as to why it is used. ececcc6
replacing full array sort with a range sort 1577ea7
Fixed null condition to ensure that null values get pushed back in the array. 76dedb5
unit tests ensure broken comparators don't put null to the start of the set and than an empty set is never sorted 5b6a2d9

The Authors

4 Egon Willighagen
4 John May
1 Stephan Beisken

Note that I always have a few patches based on doing the actual release, like setting proper version numbers. Thus, this release is really courtesy to the EBI developers.

The Reviewers

4 Egon Willighagen
2 John May

CDK 1.4.16: the changes, the authors, and the reviewers

2013-03-23T14:54:00.001+01:00

Yes, quite overdue, but here's the changelog of CDK 1.4.16. This release is about bug fixing. The first CDK 1.4 release is long behind us, and the CDK project is working hard on stabilizing the 'master' branch for the first beta releases. Because this release introduced a nasty bug, you are not supposed to use this version, but the newer 1.4.17 version instead!

The list is quite long, and includes a mix of fixes (also of unit tests), improvements, and better documentation. Things to look out for include a distribution that includes the unit tests, important fixes in comparators (affecting canonical SMILES generation), used in sorting (mind the note above, and use 1.4.17!), and a fix in the output length the ProtonTotalPartialChargeDescriptor descriptor. As always, you are recommended to upgrade, but skip this version, and go immediately for 1.4.17.

The Changes

Now also has a task to create a source distrib that includes the unit tests (per user request) 26384a5
Expanded explanation of Sigma Electronegativity Descriptor 3f02b74
Added missing literature reference, and minimal additional information requested in bug #1285 4cf0f31
Correct unit test name and added missing unit tests and annotations. 601d0af
Added unit test and annotation for moment generation (3D Similarity) 33c99d1
Added explicit atom typing and aromaticity detection to several failing fingerprinter tests. ff8d374
Changed saturation conditional to also check that a bond is not aromatic. f37cfa7
Throw an error when a non-ring is attempted to be closed 9c4bd5e
C1C1 is an invalid SMILES; expect an thrown error fc63102
Properly pad the resulting descriptor value list to MAX_PROTON_COUNT, fixing two failing unit tests 7627dhttp://1.4.1784
Small fixes: remove STDOUT use and proper int-based assertEquals() method 7a81ea3
Typed the input class and added support for IChemFile.class input 51cf991
Added a missing dependency on the test data 76ade35
Added a missing dependency 2e6fc94
Properly implemented Isotope.compare() for data and silent 09868f4
And the Groovy deps for the CDKSourceCodeWriterTest 2e62bde
Also updated the Eclipse .classpath 3f88256
Also updated the other two PMD config files for 5.0.1 fa99bc8
Updated PMD to version 5.0.1 44572c6
Updated comparator on AtomPlacer. The comparator now uses the built in Integer comparator. To avoid null pointer exceptions the access to the weight value is wrapped in a private method which return the minimum integer value when the weight is not set (i.e. null). 0d658df
Corrected null handling on AtomContainer 2D centre comparison. Previous if either container was null the comparator would incorrectly return '0' indicating they were equal. The method was changed to always provide a minimum Point2d when an atom container is null. This will sort all null containers lower then non-null containers. ed399df
Replaced subtraction based comparator. This comparator is unlikely to overflow but the safety of using the equalities ensures the proper behaviour of this comparator. e5b1bc6
Replaced subtraction based comparison. When the difference (which is a double) is converted to an integer the result may overflow. This overflow can occur in rare cases but would cause the comparator to be non-transitive. 425c795
OK, simple fix: reset the test graph before each test method call 40536af
Make sure the tests are also run with JUnit4 6e8533a
Corrected canonical label sorting in the SmilesGenerator. This canonical labels are long and thus when the difference between two values is taken the result may be larger then the largest possible integer. This in rare cases cause the value to overflow and thus make the comparator non-transitive. 704280a
corrected handling of null value in the TreeNodeComparator. Previous if any node or atom were null the comparator would return '0' which is incorrect. If one is null and the other is not the objects are not equal. d7a55d3
Generified comparators, removing redundant casts 5a5db14
Corrected comparator in InvPair sorting. The current comparison was in violation of the comparable contract which was throwing an IllegalArgumentException on JRE 7. 2f92c91
A few unit tests uncovering some issues with == comparisons 46d48f9
Added accessor for the bond length in StructureDiagramGenerator 75b5fe6
Updated the @cdk.bugs taglet to support the new SourceForge project pages fbebbe0
Replaced IBond.Order.ordinal() usages with IBond.Order.numeric() 427fac6
Added a field to the bond order enumeration allow access to the numeric value for that order. 8c34bfe
unit test 7d5ce83
unit test 5bc24fa
unit test for bug 1269 (atom... not placed by SDG, causing NPE), remove catch statements, set expected e3c3590

The Authors

21 Egon Willighagen
14 John May
4 Ralf Stephan

The Reviewers

12 Egon Willighagen
12 John May
1 Gilleain Torrance

New Paper: "Applications of the InChI in cheminformatics with the CDK and Bioclipse"

2013-03-19T15:25:00.002+01:00

Last week, Ola, Sam Adams, Arvid, and I published a paper (doi:10.1186/1758-2946-5-14) on the InChI functionality in the Bioclipse, which uses Sam's JNI-InChI and the Chemistry Development Kit underneath.

This paper partly describes the earlier work by Sam on JNI-InChI itself and the integration into the CDK, but also includes the recent support for CDK's IStereoElement, OSGi bundles for JNI-InChI by Arvid, and a few new applications in Bioclipse.

These applications demo what you can do with the InChI in Bioclipse. Obviously, this involves creating InChIs for any structure drawn in Bioclipse (that is old). New is that the manager now also support creating InChIs with particular layers. For example, with fixed hydrogens:

mol = cdk.fromSMILES("OC=O")
sinchi = inchi.generate(mol);
inchi = inchi.generate(mol), "FixedH");

But the more interesting bits are next. For example, the InChI is ideal for look up, and can be used in decision support with knowledge bases.

But as Christopher Southan showed in his "InChI in the wild: an assessment of InChIKey searching in Google" paper (doi:10.1186/1758-2946-5-10), the InChI is good for finding useful information on the web. I have taken a different approach with Isbjørn, which does not use Google, but Linked Data approaches to find information on the web. This semantic search is seeded with the InChI.

The third examples exposes work done by Mark Rijnbeek, formerly in the group of Christoph Steinbeck, who implemented a method that uses the InChI library for tautomer generation for the CDK. This functionality is now exposed in Bioclipse too. Obviously, this functionality is limited by those of the InChI library to generate those tautomers. But if you like to try it, you can do this with:

// no aromatic rings that make it hard to
// see where the double bonds are
jcpglobal.setShowAromaticity(false);

inputSMILES = "c1ccccc1O";
inputName = "phenol";
inchi.generate(
cdk.fromSMILES(inputSMILES)
)
tautomers = cdk.getTautomers(
cdk.fromSMILES(inputSMILES)
)

file = "/Virtual/" + inputName + ".sdf";
cdk.saveSDFile(file, tautomers);
ui.open(file);

Details on how to try all this in practice can be found on this page. And I am looking forward to hearing what you think of it, how you like to use it or are using it. If you like to extend it, the source code is on GitHub.

Spjuth, O.; Berg, A.; Adams, S.; Willighagen, E. Journal of Cheminformatics 2013, 5, 14+.

Source Code documentation: what JavaDoc should look like, and why.

2013-03-05T10:57:00.002+01:00

Over the years, I have blogged quite a bit about JavaDoc. JavaDoc is the system commonly used to annotate Java source code for human oriented documentation, and complement source code comments. I will not attempt to summarize what I wrote in the past, but will give some links. The primary user of the JavaDoc is, of course, the user:

And we take effort in producing good JavaDoc (despite it being the unpaid, least favorite task):

And we use tools here to support us:

And using JavaDoc taglets and doclets, we enrich the regular JavaDoc with more detailed information:

The CDK uses the JavaDoc not just for documentation, but also for compiler instructions:

CDK Module dependencies #3

Business card with ORCID and Twitter account

2013-02-28T08:00:00.002+01:00

Since I do not know how long, I am again using business cards, very useful at the QualityNano meeting I am now. There is a fixed Maastricht University template, with little room for customizations (or so it seems, at least), so no nice QR code pointing to my home page, but I managed to reuse a "title" field to list my twitter account and my ORCID!

Attending the QNano^WQualityNano meeting in Prague

2013-02-27T16:20:00.001+01:00

I flew in this morning and now attending the QualityNano (until recently called QNano, see their announcement on the right) conference in Prague. The meeting attracted over 100 people (probably closer to 200) working in the field of safety of nanomaterials, and the talks are of quite good quality and rather interesting so far. There are many, many posters too, among two of mine. One shows my nanoQSAR work the other showing the Open PHACTS examplars.

Many of the talks so far study or involve the inflammatory pathway as route of response to exposure to nanomaterials. It seems WikiPathways has several related pathways, like these two (using the Pathway Widget; both for human):

Inflammatory Response Pathway

Interleukin-11 Signaling Pathway

Several new pathway alternatives were presented, among one signaling pathway which requires a cellular bilayer to function; a single layer of cells does not make the signal possible at all.

(For the non-techies among my readers, ^W is the shortcut in many editors for deleting the left-side word seen from the cursor.)