Ryan's Blog on NMR Software

Most Accessed Article on Journal of Cheminformatics of All Time!

An article co-authored by several ACD/Labs scientists, the Russian Academy of Scientists, and ChemZoo, Inc. has been generating a lot of interest over the past year.

In fact, this article entitled, "Computer-assisted methods for molecular structure elucidation: realizing a spectroscopist's dream" is the most accessed article of all time, on the Journal of Cheminformatics website.

The news gets better. This article is FREE for your reading pleasure! As per the policy of the Journal:

"The Journal of Cheminformatics offers scientists the opportunity to publish their research rapidly in an open access medium that is freely available online to researchers worldwide. All manuscripts submitted to the journal are subject to rigorous peer review. Published material may include electronic supplementary material, such as animations (e.g. graphical chemical structures) or data (e.g. spectra) that can be downloaded for a reader's use.

The Journal of Cheminformatics aims to work closely with operators of open data repositories such as PubChem. Requests for collaboration are welcome; please contact the Editorial Office."

This motivation for this article was to mark the 40 year anniversary of published developments regarding CASE (computer assisted structure elucidation systems).

It a very interesting read that draws from many meaningful publications on CASE over the last 40 years and tracks history to outline some of the approaches that have been considered, implemented, and successful.

Some of the more significant developments such as fuzzy generation, used to deal with the ambiguity of 2D NMR experiments, as well as the determination of relative stereochemistry of large and rigid molecules with many stereocenters are discussed in this article.

I could spend more time talking about this article, but as I've already mentioned, it's FREE and can be accessed here:

http://www.jcheminf.com/content/1/1/3

Enjoy and feel free to leave any comments for the authors, and I will be sure to get them. 

How do Medicinal and Synthetic Chemists use NMR Software?- Part 6

This is a continuation of a series of posts that began here.

In the previous post of this series, I discussed how chemists can extract multiplet reports in journal format on the fly using the same procedure to re-cut integrals in the spectrum. Well in addition to having this nice reporting tool, the software can make valuable use out of this multiplicity information.

The software has the ability to assist the chemist with the interpretation of their spectrum. It does this by using the NMR prediction algorithms embedded in the software. With the use of these algorithms structure interpretation and verification can be employed.

Essentially when a chemical structure is attached to a spectrum, the software will predict a spectrum in the background and automatically compare the predicted spectrum to the experimental spectrum. To do this automatic evaluation, the software will compare three properties between the spectra:

1) Chemical Shift
2) Integration
3) Multiplicity information

As a result, by re-cutting your integrals in the software (and hence defining multiplet patterns and coupling constants) the software will use this information to help a chemist interpret their spectrum. Specifically, it will use this information to provide the chemist with immediate feedback as to what atoms in the chemical structure a multiplet is consistent with. See the movie below:

The Balancing Act of False Postives vs. False Negatives

A few months back, I referenced Derek Lowe's excellent blog, "In the Pipeline"

Some of the most interesting entries for me are:

Backtracking, Necessary, and Unnecessary

I Can Has Ugly Molecules?

Oops.

One of the take home messages for me, is when Derek says, "False Negatives and False Positives are waiting in your dataset, depend on it"

I think this is a point that is largely acknowledged in the industry. For example, I think it is pretty obvious to most organizations that there are a significant amount of incorrect structures in their registration database. Some will have a pretty good idea of how much, and others will have no idea.

I think the key point however, is to acknowledge it.

Now this brings me to the topic of the balancing act between false positives and false negatives. If I go back to the application of validation compound registrations with automated NMR verification. This application is one I have blogged about quite a bit over the last year and a half.

The key point to emphasize here, is that this application is NOT replacing the chemists analysis and QC

Chemists are still looking at their data and registering their compounds the normal way. But is an automated validation step necessary after the fact to ensure the quality of the registration database is accurate?

I asked the question from an earlier post, "what is the acceptable limit of false positives and false negatives for automated verification by software for the evaluation of  registered compounds in a library?"

I didn't get any comments on the blog, but I got a few email responses and I have discussed this with some industry people over the last few months. Not surprisingly, the results vary significantly.

Of course everyone would prefer a 0% False Positive Rate, and a 0% False Negative Rate. Yes, in a perfect world that would be great. But it's not going to happen. Not ever. So decisions need to be made about how important this this exercise is, and furthermore what is deemed acceptable?

If we acknowledge that there are incorrect structures in the registration database, then the question is how much, and furthermore, how many can be removed with the use of some automation.

I think the challenge that most organizations face today when evaluating such a system is trying to balance out to sides, false positives vs. false negatives. Why?

My understanding is that if you have a system that is false positive tolerant, well, you are sacrificing the overall quality of your organizations registration database. On the other side of the coin, if you have a system that is false negative tolerant, well, someone needs to look at the data manually to see if this is indeed the right or wrong structure. So there are two forces pulling here. One, is actually knowing that despite your efforts, you are still letting incorrect structures in your database. The other force is knowing that you need a real person's time to spend on manually evaluating a bunch of spectra.

Most spectroscopists I know, do not have a lot of time for this. So of course it would be preferable to have a system that is potentially more false positive tolerant. However, the whole point of implementing the system in the first place was to identify the false positives.

Tough decision...

But at the end of the day, I believe there are systems and applications out there that can help drastically improve the quality of your registration databases.

However, the balancing act between false positives and false negatives is an issue that you are most certainly going to have to juggle.

Yes, Chemists Will Still Have To Look at their NMR Data

I've been away for the better part of the last few months, haven't been able to spend anytime blogging, and I apologize for those of you who have subscribed to the blog and anticipate hearing from me on a regular basis. 

During my travels I have had a lot of discussions about ACD/1D NMR Assistant our software tool that has been designed specifically in mind for synthetic and medicinal chemists. Based many of these discussions I feel like I should address this idea of structure verification algorithms in software.

I think the point is NOT to replace the chemists QC or review of their NMR analysis. I know that may be an attractive marketing message...but that's not the purpose.

A chemist runs an NMR to ensure that their reaction has come to completion, and to confirm that they indeed made what they thought they made. Of course they have other reasons for running NMR (quantitation for example) but this is generally the major reason.

I think there is widespread acknowledgment throughout the industry that the detail and emphasis put into this personal QC varies greatly from chemists to chemist. Some chemists are very diligent in their NMR analysis assigning their data and ensuring that every peak and multiplet makes sense. From my own personal experience as a graduate student and chemist in the industry, I will be honest that for every NMR I ran following a reaction, I didn't diligently assign every multiplet from every ¹H spectrum. If I was in the 3rd or 4th step of a reaction, I was simply looking primarily at 1 or 2 peaks to see if my substitution was successful. If that peak disappeared at 3.1 ppm or that peak appeared at 2.7 ppm...I was off to the next reaction.

Does that make me an irresponsible chemist? Perhaps?...But I keep hearing that this practice is pretty common in the industry. In fact, I've had some chemists admit that they don't look at the NMR at all. If the LC-MS looks good...it's time to move on.

I think that this is precisely  where a tool like 1D NMR Assistant fits in. No, it's not necessarily smarter than the average chemist, and it's not necessarily better at assigning data than a chemist. But it can be a valuable tool to complement the chemist's workflow and help them make better decisions.

Further more....it's not a black box that tells you whether something is right or wrong.

Here's an example output from the software:

Note the language:

The structure does not appear consistent with the spectrum.

It is not suggesting that the structure is wrong only that it "appears inconsistent"

This is a key point. It is flagging certain parts of this spectra that perhaps the chemist should spend an additional 10-20 seconds looking at to ensure that they indeed made what they thought they made.It does this because it takes a brute force approach towards the assignment of all protons in the chemical structure to multiplets in the experimental spectrum. If it can't find a matching multiplet, or the assignment doesn't make sense...it will tell you.

In some cases these flags may not be serious. Maybe it's a solvent peak, or a slight impurity, or the software didn't predict the chemical shift accurately. That's OK...but the purpose is to alert the chemist that this is one they might want to take a look at. One that might otherwise end up as another false positive in the organizations registration database. Every once in awhile the software will uncover something that might have gotten overlooked.

In addition to being a standard 1D NMR Processing package, it is this functionality that has made 1D NMR Assistant a popular choice early in it's first year of availability.

False Negatives and False Positives are Waiting...

Great post from Derek Lowe from In the Pipeline the other day talking about the dangers of not quality checking those fine-looking starting compounds for your project. Chemistry happens and yes, mistakes do too.

In fact, it appears that Derek has been on a kick as of late referring to personal QC.

I Can Has Ugly Molecules?

Oops.

I thought this would once again be a good opportunity to provide you with a link to a poster Sergey Golotvin presented at ENC 2008 entitled, "Validating the Quality of Large Collections of NMR Spectra Automatically".

Long story short, 15,000 1H NMR Spectra from the Aldrich collection were evaluated in complete automation and the software was able to confirm 88% of the collection as having chemical structures that were consistent with the respective spectra. In addition, 4% were flagged by the software as being inconsistent. A closer, manual look at those 4% revealed that there were indeed some truly wrong structures (or incorrect tautomers) in the collection.

This was evaluating the 1H NMR data only. Using additional 2D experiments, such as HSQC, will likely improve these results.

Just an example of a check an organization can build into their process for additional QC of their registration database for example.

Is it perfect? Absolutely not. There are perhaps a few more false positives in there that the software didn't catch, and of course the software provided some false negatives as well, annoying because presumable someone has to look over them manually only to realize that they were indeed the right structure all along. But at least this doesn't involve manually pining over 15,000 spectra!

We continue to run these datasets, and actually have a consortium consisting of several NMR experts in the industry we call ASCI (Automated Structure Confirmation Initiative) where we are testing and validating this technology in the real pharmaceutical world. Identifying the common areas where false negatives and false positives occur and trying to address them with algorithms.

Will we ever solve all the problems, especially in the world of novel chemistry? Of course not, and for that matter there are some existing problems that appear to be too hard to solve.

But that being said, what is the acceptable limit of false positives and false negatives for automated verification by software for the verification of  registered compounds in a library?

Interested in hearing your thoughts.

Automated Structure Verification by 1H NMR Only

I've blogged several times about the progress and applications of automated structure verification with the help of ACD/Labs software.

There are really two main approaches right now:

1. Combined verification which includes automatically verifying the correspondence between a proposed chemical structure and the 1D ¹H and 2D HSQC spectra. I've blogged about the publications and applications of this method previously.
2. ¹H only verification. Of course the first approach is preferable from an accuracy standpoint as the additional information gained from the HSQC spectrum increases the selectivity and specificity of the results.

However, I always get questions on how well we perform on ¹H NMR only because in some organizations and environments it is simply not feasible to always run an HSQC in tandem with a routine ¹H NMR analysis.

As mentioned in previous posts, we've already conducted a validation study on this approach and it was published in a 2006 article in MRC. We've continued to investigate and validate this approach and we recently presented our latest results using 1D ¹H NMR data only at ENC 2008.

The poster highlighted a study on the automatic evaluation of over 15,000 Aldrich compounds and spectra from Aldrich NMR Spectra Database.

The results of this study revealed that the software was able to confirm 88%of all spectra as consistent and flagged less than 5% as inconsistent.

One of the more interesting discoveries in this study was that it revealed some truly wrong structures in the Aldrich NMR database.

More information on these as well as shortcomings in prediction, processing, and analysis are provided in the poster that can be downloaded:

http://www.acdlabs.com/download/publ/2008/enc08_aldrich.pdf

What's New in Version 11- Improved 1D NMR Structure Verification Accuracy

Automated structure verification using ACD/Labs software is a method that compares the chemical shifts, intensities, and multiplicities of signals in experimental NMR spectra with those from a predicted NMR spectrum with a proposed structure.

Naturally, in order for this process to be effective the chemical shift prediction, multiplet characterization, and integration measurements must be accurate. This process is described in detail in a 2006 publication in MRC.

As mentioned in the previous post, in version 11 the automated multiplet analysis
algorithm in ACD/Labs software has been significantly improved.

The purpose of this post is to show you how much the improvements of the automated multiplet analysis algorithm in version 11 impacts the performance of automated verification of ¹H NMR spectra.

The following results compare the verification improvements in version 11 based on the multiplet analysis enhancements ONLY. In version 11 we made several other improvements in NMR prediction and analysis, and I will get to them in a future post. For now:

For this study, two different data sets each consisting of the ¹H NMR spectra of 30 samples and the correct chemical structures were automatically processed, analyzed, and evaluated by the software.

Test Set 1- A set of 30 spectra and their respective chemical structures with reasonably good signal to noise:

Test Set 2- A set of 30 spectra and their respective chemical structures with lower signal to noise:

As you can see clearly, the improvements in multiplet analysis in version 11 heavily impact the performance of automated NMR verification of ¹H NMR data in both datasets. Based on the dataset used for this study, the software (employing the version 11 multiplet analysis algorithm) was able to correctly confirm the consistency between the proposed chemical structure and the experimental spectrum ~60% of the time in both datasets.

For a real world application of this system, check out a previous post that described Anthony Macherone's workflow at ASDI (and a link to his presentation).

The next questions is, "If you use more data, how well does this system perform?"

A future post will describe a study that highlights the latest performance statistics of a "combined verification" approach that can automatically identify  correct and incorrect chemical structures based on their 1D ¹H and 2D HSQC NMR data. This system was described in a 2007 publication in MRC.

Stay Tuned.

ACD/1D NMR Assistant Part 2- Assigning NMR Data

The movie below (if you receive this by email you will have to come to the blog) highlights the unique Multiplet Assignment Preview available in ACD/1D NMR Assistant.

This feature helps users evaluate potential assignments by considering the chemical shifts, multiplet properties, and integration values of experimental multiplets in the spectrum.

Just another reason to embrace NMR software. A piece of paper can't do this:

If you can't get the above video to work, or if you want a bigger version , click here to watch the demonstration.

The above example is just for illustration purposes. The best way to evaluate the software on your own data is to go to the ACD/Labs website and request a free trial.

Stay Tuned....next up, the structure verification algorithm in ACD/1D NMR Assistant!

Comparing NMR Prediction Approaches

A new article is now available on JCIM ASAP entitled,

"Toward More Reliable ¹³C and ¹H Chemical Shift Prediction: A Systematic Comparison of Neural-Network and Least-Squares Regression Based Approaches" 10.1021/ci700256n

A very interesting read for those interested in the performance of NMR predictions.

Some really nice conclusions on both the speed and accuracy of different NMR prediction approaches using our very own ACD/CNMR Predictor.

As Tony is an author of this piece, head on over to his blog for his synopsis on the work as well as the abstract.

New Product Time! Introducing ACD/1D NMR Assistant

I apologize for not posting much lately, but things have been pretty busy in the ACD/Labs NMR world.

I have alluded to this moment in some previous posts for awhile now and I am now happy to unveil a new NMR product to be offered by ACD/Labs.

ACD/1D NMR Assistant.

So first and foremost, what is it?

I have talked a lot about how synthetic chemists currently use ACD/Labs software and blogged about the benefits and key features in the software.

The bottom line for us, was that while we have been successful in selling
ACD/1D and 2D NMR Processor to chemists and students in industry and academic institutions, we believe that there was still quite a bit of work to be done to design a tool for the chemist.

The truth is, ACD/1D NMR Processor has been around for quite some time. From the very beginning, we were developing a product with the NMR Spectroscopist in mind. Naturally, during the evolution of this product many sophisticated and advanced features have been added and as a result the software can sometimes be viewed as bloated, and overly complex for a novice or non-expert user.

In recent years we have worked hard on continuing to add advanced features but to also try and simplify things. At the end of the day, we made the decision to go in a different direction and create a separate product. This way we can continue to develop ACD/1D NMR Processor with the Spectroscopist in mind, and build ACD/1D NMR Assistant with the synthetic chemist in mind.

So as a result, we developed ACD/1D NMR Assistant and it is now finished and available.

How is ACD/1D NMR Assistant different than ACD/1D NMR Processor?

1. Ease of Use- ACD/1D NMR Assistant includes all of the features available in ACD/1D NMR Processor, we have just de-emphasized some of those features in the software in an effort to greatly simplify the toolbars and interface. I believe that we reduced the learning curve significantly. One example is that upon file import an FID will automatically get FT'ed, phase corrected, and baseline corrected. In addition, the software will look for solvent and water signal automatically and darken them out. Because it includes Shortcut Mode from NMR Processor, users can peak pick, integrate, and characterize multiplets with one simple click and drag over each multiplet.
2. Assignment Assistance- The big improvement added to ACD/1D NMR Assistant is that the software can now provide the user with feedback on potential assignments. When a structure is proposed users can hover over a multiplet of interest and the software will provide real-time feedback as to what the best assignments are. It does so by highlighting atoms in a structure with a green, yellow, and red color scheme.
3. Structure Verification- Whether a user simply wants to check their own assignments, or ask the software to provide feedback on the consistency between a proposed structure and an experimental spectrum, the software provides this capability via one simple button click. If the software deems a spectrum-structure pair as inconsistent, it provides direct feedback on which part (or parts) of the spectra should be looked at closer and what the specific problem is.
   

The ultimate motivation behind this product was the simple idea that, "an NMR spectrum is a means to an end"

While there are several software packages that can process NMR data and print out spectra for evaluation, there is nothing available that really helps users evaluate and interpret an NMR spectrum  and it's relationship with a chemical structure.

That is until now, IMHO.

I think the development of ACD/1D NMR Assistant changes all that.

One more thing...

This product wasn't just built by a few software-oriented people within ACD/Labs. We have been speaking with current customers and chemists for years on how we can improve our offering. I personally have spent a very considerable portion of the last 2 years speaking with chemists, students, and spectroscopists who support chemists, discussing what was lacking in our current offerings and how we can improve our software for these users. Further, I spoke with people who have never used ACD/Labs software and discussed their desires and expectations for an NMR package that would suit their needs.

Finally, after the first stage of new development we sent the software to groups of chemists in 5 different pharmaceutical organizations in the US and had them evaluate the software. These groups ranged from chemists who currently used our 1D NMR Processor, to chemists who had never used it. In doing this we hoped we would be able to appropriately gauge the acceptance and learning curve of the new software. This wasn't beta testing...this was MARKET testing. We are very grateful to these chemists who provided valuable feedback that was implemented in the final version of the software. They were able to point out some very obvious things that our software-focused minds simply overlooked for years.

Following these evaluations, we took the feedback we received and performed another round of development + evaluation to optimize our offering.

And through those exercises, we reach today. A finished product in hand that I am very excited about.

Over the next few weeks, I will be highlighting different features and workflows in the software to educate you on how it works and what it does.

For those of you who want a sneak peek right now, go over to the ACD/Labs website and view the movie that shows the software in action. If you like what you see, just fill out the form at the bottom of the page and you can get a free evaluation copy to try yourself.

Click here to go to the 1D NMR Assistant page.

I hope you like it!