Depth-First - Latest Comments

Re: Thirty-Two Free Chemistry Databases

Doug — Mon, 13 Feb 2012 14:26:22 -0000

Hello,

Another very useful database is:

Synthetic Focus

http://www.syntheticfocus.com

Re: George Whitesides: The Concept of the Scientific Paper is Eroding Before Our Very Eyes | Depth-First

MikeyMedChem — Mon, 30 Jan 2012 13:36:55 -0000

This is all very true, and I am excited at the prosepects. Scientists need to change the practices to adapt to the new realities. However, academic scientists will do those things that are rewarded (read: tenure decisions and positive grant reviews driven by reviewers who simply count publications and impact factors as a measure of productivity). Until that bean counting mentality behavior is discouraged and corrected, I don't expect a lot of change in the direction that Prof Whitesides is suggesting.

Re: IBM Donates Large Collection of Patent Chemical Structures to NIH/PubChem | Depth-First

Rich Apodaca — Sat, 28 Jan 2012 16:34:29 -0000

See also this newer presentation from Stephen Boyer.

Re: Five Things to Do Instead of Protesting the Research Works Act (HR 3699) | Depth-First

Rich Apodaca — Thu, 26 Jan 2012 12:01:04 -0000

For those who think that taking a stand in chemistry is impossible, consider Mat Todd, professor of organic chemistry at the University of Sydney, who will no longer be publishing for or reviewing for Elsevier journals. This includes Tetrahedron and Tetrahedron Letters.

Re: Five Things to Do Instead of Protesting the Research Works Act (HR 3699) | Depth-First

William Gunn — Wed, 25 Jan 2012 13:56:05 -0000

I've done 3 out of the 5, but got no response. There are some big name scientists, like Jonathan Eisen and Phil Bourne, who have had the courage to take a principled stand, but OA advocacy doesn't have nearly the budget that the big publishers do, so just reaching people is hard. Furthermore, that addresses only one leg of the three parties who have to make a move. It's great to advocate pushing on scientists, but they'll tell you that as long as their career success depends on being in a big name journal, their hands are tied, so it's smart to advocate trying to reach senior scientists, but having worked for such a fellow, I despair of that being successful. I think we're going to have to wait until some of the younger guys become more senior and can afford to rock the boat before we see movement there, which leaves us with publishers and librarians. Many librarians are quite vocal [1], but their hands are tied by faculty. There are two options for movement on the part of publishers, new publishing models, like the one PLoS ONE uses (quite successfully[2]) or getting the federal government to put pressure on the publishers to do the right thing via a Public Access Policy. Mandates work. [3]

I think we should be doing ALL these things. It's not a zero-sum game. Having one group that is so motivated push on boycotts and writing scientists while another differently motivated group advocates for a public access policy doesn't harm overall progress. I'm just as annoyed with my colleagues for being apathetic and I appreciate what you're trying to do, but why take the zero-sum approach?

1. http://www.cdlib.org/cdlinfo/2...
2. http://www.michaeleisen.org/bl...
3. http://poynder.blogspot.com/20...

Re: Two Chemistry Apps for Mobile Devices - Depth-First

yalda — Wed, 25 Jan 2012 01:17:01 -0000

These apps are pretty good!

You can also try the following website: www.apolloscientific.com/iphon...

It has some very very useful and helpful tools for all chemists/scientists.

Good luck:)

Re: Five Things to Do Instead of Protesting the Research Works Act (HR 3699) | Depth-First

Rich Apodaca — Tue, 24 Jan 2012 18:44:46 -0000

William, thanks for the great feedback! I'll try to respond the the main points, but this will probably take a post of its own to really digest.

"in cases
where there's a clear public need, but none of the stakeholders want to
make the first move, having the government step in and make something happen is an option we need to keep on the table. "
You have a point. My concern is that the Public Access Policy is the wrong approach. Your phrase "screwed up detente between scientists, publishers, and librarians" rings true. And government can be effective at breaking some logjams.

I disagree that the Public Access Policy is the way to do this. I'll go further and say this specific law is counterproductive. It does nothing to actually change the behavior of scientists or libraries, and only slightly forces behavior change in publishers. It allows all parties to continue with business as usual and so keeps the status quo intact.

"Another option would be for the federal government to prevent federal funds going to libraries from being spent on subscriptions to journals which don't give authors redistribution rights."

Now, this is an idea that has a chance of doing what we both want. The problem is that the old-guard publishers will fight any bill like this and win because there's no money (or popular support) on the other side. This is the problem with any legislative approach, and it's why I oppose them in general.

Old-guard publishers draw their power ultimately from the prestige they built up prior to the digital age. They are now abusing that power to retard scientific progress. We as individual scientists are the only ones who can take this power from them.

Call me naive, but there must be a few high-profile scientists out there with the courage to step out of line and call bullshit on a corrupt system that's harming science. It would only take a few to start a long-overdue discussion and break the logjam.

I'm curious - have you ever tried any of the 5 suggestions I offered? If so, what were the responses?

Re: Standardization in Lab Automation: The SiLA Consortium | Depth-First

Jwaters — Tue, 24 Jan 2012 16:42:55 -0000

Great article! I really enjoyed it!

http://www.hudsonrobotics.com/

Re: Five Things to Do Instead of Protesting the Research Works Act (HR 3699) | Depth-First

William Gunn — Tue, 24 Jan 2012 15:58:20 -0000

How about in addition to protesting? What we have right now is some kind of screwed up detente between scientists, publishers, and librarians. No one is willing to make the first move. Scientists won't change until grant/tenure review changes, publishers won't move as long as they can continue to rake in 30% profit margins with essentially no competition (although, see PLoS ONE) and librarians don't have the desire (or ability in some cases) to just stop subscribing to the worst offenders.

I'm not generally in favor of government intervention, but in cases
where there's a clear public need, but none of the stakeholders want to
make the first move, having the government step in and make something happen is an option we need to keep on the table. Another option would be for the federal government to prevent federal funds going to libraries from being spent on subscriptions to journals which don't give authors redistribution rights. I think we should be doing ALL these things.

I agree we shouldn't blame the publishers for acting in their rational self-interest, but unless we want to wait until PLoS ONE style journals take over (which is likely, but not guaranteed) we need help from the government. When you contrast the amount of lobbying and wining and dining of rockstar scientists that the publishing industry can do with the amount that an organization like SPARC can do, it's no wonder things aren't progressing very fast. I'm not turning down any help we can get, even suboptimal solutions like Pubmed Central. It's far easier to transition from having everything in Pubmed Central to having everything online and OA from many sources than it is to go from no public access at all.

And the evidence shows that mandates work. I don't have to post the link because I know the Harnad-bot will be in here shortly with links galore, but they work. Even the limited content available there helps us accumulate data and use cases which helps non-convert scientists realize the value of such an archive. If you remember, one of the older arguments the publishing lobby used to make was that everyone already has all the access they need. They claimed that everyone who needed a paper already had some way of getting it. PMC serves 500K pageviews A DAY, 60% of which goes to non-institutional domains. The publishers don't make that argument anymore. Now they're harrumphing about government involvement and killing jobs and such. Expanding NIH policy, even as imperfect as it is, to the 11 other federal agencies and dropping the 12 month embargo would show that the jobs argument doesn't hold water either. After that, we'll be on to some other argument about why the publisher's 30% profits need government protection.

I have no illusions of this being a fast process. Recording artists have been far more innovative at coming up with new business models, yet they're still dealing with nonsense like SOPA and PIPA and the DMCA. If I were king of America, my public access to research decree probably wouldn't look much like the NIH policy, but a flawed policy is better than no policy at all.

Re: Open Source Name To Structure Conversion of IUPAC Nomenclature - OPSIN Revisited | Depth-First

Maryam1984j — Mon, 23 Jan 2012 15:40:12 -0000

2-naphtol

Re: On the (F)utility of Extending the Molfile Format | Depth-First

Rich Apodaca — Sun, 22 Jan 2012 11:07:10 -0000

Egon,

"Another point indeed is, is getting the electron counts right. At an atomic level, the CDK atom types take care of that. The CDK interface has a field for electron count too, but that is hardly used. "

That could be a problem because algorithms that depend on, for example, formal charge - where that charge is never actually calculated from electron count but rather assigned in a top-down fashion (or when the charge is non-integer) could lead to problems when taken with multicenter-bonding approaches. Benzene radical cation, for example.

I don't think it's enough for a datastructure to merely have an electron count field. All algorithms that perform interpretation of any kind as to charge or bonding need to use those fields (or derivative fields that change in response to changing electron count).

Re: On the (F)utility of Extending the Molfile Format | Depth-First

Rich Apodaca — Sun, 22 Jan 2012 11:00:23 -0000

Alex,

"You will get some pretty weird/broken results when you start tinkering with the valence & query-H-count fields in V2000 molfiles and feeding them into legacy software. "

Any examples of software/file combinations that show the effect?

"Providing a bridge across the gap, by using backward-compatible extensions, is a good way to encourage this."

Yes, that's the sense I got from your paper.

"Have to disagree with the claim that bond orders are implied by atom properties - there is a whole category of molecule formats that are broken because they are based on this falsehood."

I don't remember making that claim. Bond orders have a strict mathematical definition based on number of edges involved and the electron count of the bonding system. See: http://dx.doi.org/10.1021/ci00... But I'm jumping the gun a little...

Re: On the (F)utility of Extending the Molfile Format | Depth-First

Egon Willighagen — Sat, 21 Jan 2012 07:40:10 -0000

I tend to agree both with Rich and Alex: I strongly belief there is enough room for multiple representations, and indeed needed for the science we do. It's hard to find a one-fits-all solution, and the one we got is too complex to give results in reasonable time (QC). Cheminformatics, practically, has to deal with a number of applications, visualization being one of them, computation another; these are not necessarily compatible, though we must thrive to have clear protocols for transforming one into another. That's what the CDK is trying anyway, and certainly what I am trying to show people...

The CDK data model allows for the common chemical graph approach, where bonds have a variable amount of electrons, and also supports multi-atom bonds. I have shown previously how benzene and ferrocene can be done in various representations, in fact, also in reply to a blog by you, Rich :)

http://chem-bla-ics.blogspot.c...

But while the data model is one thing, there is also the large amount of algorithms to operate on it. As noted before, I filed a bug for the valency information not always being read into the data model.

Another point indeed is, is getting the electron counts right. At an atomic level, the CDK atom types take care of that. The CDK interface has a field for electron count too, but that is hardly used. I have just uploaded a patch to set electron counts when a bond order is set (note that the CDK does not have an 'aromatic' bond order, deliberately):

https://sourceforge.net/tracke...

Re: Digital Destruction in Scientific Publishing: Why This Scientist Supports the Research Works Act (HR 3699) | Depth-First

Rich Apodaca — Wed, 18 Jan 2012 19:27:25 -0000

Hello John,

I noticed that. This quote, taken from the linked text, might explain it:

"NPG and Digital Science exist to support the creation and dissemination of human knowledge on a sustainable commercial basis. We seek to enable the open exchange of ideas, especially in scientific communities, in line with the requirements and objectives of relevant stakeholders."

Most publisher/societies have similar mission statements if they would care to refer to them. It's good to see one publisher who actually does, and who can reconcile that mission with the imperative of paying the bills.

Re: Digital Destruction in Scientific Publishing: Why This Scientist Supports the Research Works Act (HR 3699) | Depth-First

John Spevacek — Wed, 18 Jan 2012 15:34:46 -0000

Very strange, because a bastion of imprimatur, Nature, does not approve of the Act.
http://www.nature.com/press_re...

Re: On the (F)utility of Extending the Molfile Format | Depth-First

Egon Willighagen — Sat, 14 Jan 2012 11:41:52 -0000

Rich, thanx for the 'bug' report... I have filed it here:

https://sourceforge.net/tracke...

Re: On the (F)utility of Extending the Molfile Format | Depth-First

daen — Fri, 13 Jan 2012 15:28:35 -0000

A very interesting article, thanks! I wrote about some of MDL's tools in June 2005, and it's interesting to see MOL format still hanging in there. I imagine that the past and future investments of effort in software will ensure its usage for a while. It also jogged my memory that another feature that the V3000 format introduced, which actually was quite useful, was enhanced stereochemistry representations (unknown, known relative and known absolute stereocentres).

[Edit: Apologies, I see that you already mentioned the stereochemistry improvements ...!]

Re: On the (F)utility of Extending the Molfile Format | Depth-First

Alex M. Clark — Wed, 11 Jan 2012 16:33:41 -0000

You will get some pretty weird/broken results when you start tinkering with the valence & query-H-count fields in V2000 molfiles and feeding them into legacy software. That's why I argued in favour of using M-block extensions, which are fairly consistently ignored or passed through by the thousands of custom-written algorithms currently in use.

That being said, few things would make me happier than seeing the community abandon the format altogether, and move to something well designed. Providing a bridge across the gap, by using backward-compatible extensions, is a good way to encourage this.

Have to disagree with the claim that bond orders are implied by atom properties - there is a whole category of molecule formats that are broken because they are based on this falsehood. Bond order localisation is important, with the non-Hueckel conjugated rings (e.g. cyclooctatetraene) being a prime example, or various resonance forms that have several compatible energy wells (i.e. double bond/radical placements - quinone, allyl radical, etc.) that do not necessarily interconvert at room temperature.

Re: IBM Donates Large Collection of Patent Chemical Structures to NIH/PubChem | Depth-First

Lead optimisation — Fri, 30 Dec 2011 19:01:42 -0000

Interesting read!

Re: Fast Substructure Search Using Open Source Tools Part 2: Fingerprint Screen With SQL

Yuriy Sushko — Fri, 25 Nov 2011 10:14:56 -0000

That's a very interesting reading, thank you for the link. Its is quite recent also - looks like optimising substructure search performance has some interest nowadays. The publication I mentioned is also only several weeks old.

The idea in either case is to avoid linear scan over the whole database via some smart indexing of fingerprints. I am eager to try this k-d tree approach from Tripod.

Re: Fast Substructure Search Using Open Source Tools Part 2: Fingerprint Screen With SQL

Rich Apodaca — Wed, 23 Nov 2011 17:03:56 -0000

Yuriy,

For even better substructure search performance on massive collections, ditch the SQL database altogether and use an in-memory index.

Now that's hardcore :).

For more modest databases of <=100,000 records, the approach described here works generally well in practice. MySQL supports the method outlined here, which keeps the linear scan in SQL. Apparently, Postgress supports bitwise operations on a blob field directly as one of the commenters below pointed out.

Re: Fast Substructure Search Using Open Source Tools Part 2: Fingerprint Screen With SQL

Yuriy Sushko — Wed, 23 Nov 2011 12:05:11 -0000

It seems that SQL queries with bit checking conditions (e.g., ".. WHERE FP11 & 4 = 1 ..") are not possible to execute using any standard database indices. Therefore, such queries can be quite slow on millions of compounds. Is there any way to speed up such queries?

For example, I have read a publication from Johnson&Johnson - they implemented custom domain specific indices for Oracle. This is a really hardcore way..

Re: Sixty-Four Free Chemistry Databases | Depth-First

Matteo Floris — Mon, 14 Nov 2011 03:10:17 -0000

http://mms.dsfarm.unipd.it/MMs...

Re: Sixty-Four Free Chemistry Databases | Depth-First

Rich Apodaca — Sun, 13 Nov 2011 13:27:36 -0000

Also see SCRIPDB, structure-searchable database from the patent literature:

http://dx.doi.org/10.1093/nar/...

Re: Sixty-Four Free Chemistry Databases | Depth-First

Michel02 — Fri, 11 Nov 2011 08:12:54 -0000

Hmm.. maybe jhetchem.com should be added to this list as well.