LC/MS Applications in Biomedical Research Symposium at NIH

Today was the Agilent-sponsored symposium on LC/MS Applications at the NIH.  This was easily the best manufacturer-sponsored event I've ever attended.  Normally I feel these things are big commercials for new equipment with one or two good talks in between.  This was the opposite.  I feel like I saw a number of really good talks and with limited high pressure sales techniques.  I really commend Agilent for putting on such a nice event.
Overall impressions:
1.)  The NIH main campus is a confusing place to get around.  But I made it, and I will be less intimidated about going there again.  Actually, this had nothing to do with the talk, but was a big part of my day.
2.)  I finally got to see Gary Siuzdak.  As the author of the second best Introduction to Mass Spectrometry Book, and the book that got me started in the field, I'd been looking forward to this talk since I heard about it.  He didn't disappoint. They are doing some really interesting things at Scripts, including some SELDI-like analysis of cancer cells.  A great talk.
3.)  My favorite talk, content-wise, was by a Postdoc at Johns Hopkins by the name of Patrick Shaw.  He showed some really interesting SILAC-based work using the strangest cancer cell lines I have ever heard of.  The highlight of the work was in the way he implemented SILAC labelling with subcellular fractionation to implicate integrin translocation in tumorigenesis.  Really smart work.
4.) Another good talk came out of Hopkins from Dr. David Colquhoun.  While I still don't understand the premise of his talk, linking malaria and HIV, I was impressed by the scope of the work.  They performed proteomic analyses on the membranes from one particular sub-organ of mosquitoes.  This involved removing the midgut from an estimated 30,000 mosquitoes.  This is a task that only a postdoc at Hopkins would perform.  I can NOT wait to dig up their data from the repository.

All in all a nice day where I learned a lot and got to walk across a nice campus on a very spring-like day.

Proteomics in Space and Time (PROSPECTS)

Molecular and Cellular Proteomics this month will focus on the PROSPECTS project.  If you are not familiar with this initiative, it is a collaboration between several of the top proteomics labs in Europe and some of our favorite equipment manufacturers.  The goal of the project is to create second-generation proteomics methods, instruments, and techniques.  We are all probably familiar with some of the results of this collaboration, as the initiative has already generated >80 publications.

Surprisingly, however, with as much as the published results have pushed our field further toward true -omics, the best work appears to be on its way.  Even if you only have a passing interest in where proteomics is going in the future, you should read this early release description of the project.  Of the many revolutionary projects this group has begun, the one of particular interest to this researcher is the use of real-time peptide sequencing! I will continue to detail these reports as they become available, and as I get the opportunity to wrap my mind around them.

Peptide Validator vs. Percolator, Part 3

As I mentioned a few days ago, I am trying to get to the bottom of this Percolator module added to Proteome Discoverer in version 1.3.
I finally got around to the original citation: "Semi-supervised learning for peptide identification from shotgun proteomics datasets."
The paper begins by describing how normal false-discovery rates (FDR) are used in proteomics.  The FDR is applied after all the peptides have been scored.  If the FDR is set at 5%, then the lowest scoring 5% of peptides are dropped for being artificial.
The authors note that this can be a problem because those peptides may, in fact, be real, or you may be letting through bad peptides all based off of an arbitrary cutoff.
Percolator works by creating a decoy database made up of scrambled peptide sequences, presumably created from the FASTA file you are using which are used as negative examples.  The best scoring peptides are used as positive examples.  Percolator uses these examples to train a "machine learning algorithm."
The paper goes on to show how Percolator works better than XCorr plus a FDR, by evaluating a tryptic digest using Sequest with no enzyme selected.
All in all, they provide a pretty convincing argument that this works well.

Problem:  In their analysis, the Sequest algorithm took 3 days to process their sample and Percolator analysis required only 4 addition minutes.

As noted in my previous entry, the absolute time limiting step in my analyses is the Percolator module.  On single runs, Percolator takes 2 - 2.5 times as long as Peptide validator.  On a complex data set, Percolator has added as much as 2 days to my processing.  I don't know why this module takes so much longer than the one described in this paper, but until this is resolved, high throughput labs like mine are going to find it too much of a handicap to employ.

Global kinomic and phospho-proteomic analyses of human malaria parasite Plasmodium falciparum -- new favorite paper

This extensive (almost exhaustive) paper has been my free time reading for the last day and a half.  I was aware that some grants had been funded looking at the phosphoproteomics of the malarial parasites and have been eagerly awaiting the results.  I had no idea that something this good was in the pipeline.

Phosphopeptide enrichment was performed in a unique way:  Peptides were separated by hydrophilic interaction liquid chromatography (HILIC) and the fractions were first enriched by IMAC, and the flowthrough was enriched by TiO2.  I have always performed TiO2 first, then IMAC, but I see no reason why we wouldn't go the other direction with it.  If anything, the propensity for IMAC to enrich acetylated peptides might mean that the TiO2 enrichment may be enhanced.
The output was ~1,200 individual phosphorylation sites, primarily from the parasite. The best part?  They found tyrosine phosphorylation sites.  The researchers were obviously as surprised as I was, as the primary take away from both Figures 3 and 4 were that they were true phosphotyrosine sites.  This is a big deal because there are no predicted tyrosine kinases in the parasite.  Yet these critical, fine-tuning phosphorylation events are occurring somehow.

The phosphoproteomics is not, however, the most impressive part of this very nice paper (although the timing was really good, considering we were about to start it ourselves!!).  The best part is that they went through the organism and knocked out every gene that looked like a kinase in the parasite.  By doing this they were able to determine what kinases were essential AND when.  I will be referring to every table in this paper on a weekly basis, as our data is processed.  I love when you see a paper in Nature or Science and have no doubt that it belongs there.

You can find this incredible work here.

Thermo Application Note: 557; The Elite is Better than your Orbitrap

Our sales representative from Thermo-Fisher sent out this new application note (557) yesterday.
The summary is this:  "The Orbitrap Elite is better than any MS device ever made, period."

In this tech note, authors Reiko Kiyonami, Martin Zeller, and Vlad Zabrouskov from Thermo show the superiority of the Elite in performing high resolution, accurate-mass mass spectrometry (HR/AM) (mass mass?)
This is the same procedure I outline in my Methods book, that I refer to as HRSIM, or high resolution single ion monitoring.

Unsurprisingly, the Elite excels at this method, whatever you call it, considering that it has a resolution 2.5 times better than the Velos, and performs 60,000 FWHM scans at 4 times the speed of the Velos, it is definitely a step forward.

The bigger question for me is:  has the ion trap gotten any better at filtering?  Anyone used to doing MRM experiments that has moved to HRSIM on an Orbitrap has been disappointed by the fact that the LTQ on the front end has a mass window no smaller than 1 Da.  One day they'll put a quadrupole trap on the front of the Orbi and blow all of our minds.  But until that day, the Elite is definitely the best thing out there.

Lunch time reading: Systematic and quantitative comparison of digest efficiency and specificity reveals the impact of trypsin quality on MS-based proteomics

The newest issue of JPR contained the following article, you can read the abstract here.
The title is the lengthy:  Systematic and quantitative comparison of digest efficiency and specificity reveals the impact of trypsin quality on MS-based proteomics.


I think it does a good job of addressing a question many scientists have considered -- why am I paying so much for sequencing grade trypsin, when trypsin is cheap in bulk.
The very thoroughly explored answer is in this paper.  Lower quality (cheaper) trypsin = lower quality MS results.  They examine the 6 different types of trypsin (and aren't afraid to list the manufacturer or distributor) and show what is good (which is fortunately what we buy!!!) and what leads to unassigned fragmentation spectra.

Percolator vs Peptide Validator Part 2

So I finally had some free time to really look at Percolator vs. trusty old peptide validator.  This has been due to the fact that our Processing PC lost 3 hard drives (manufacturing defect, or my work simply kills PCs).  Since the next fasted PC I have access to is my home computer we built this winter, I got an emergency license from Thermo and loaded it at home.  Our PC is back with 2 new hard drives, but I'm still using my home PC to get caught up on the enormous backlog that we've generated.

Here is the experiment:  
1.) A single fraction (SCX separated) of mouse serum was ran on the Velos on a standard 140 min gradient using the Top10 method.
2.) This fraction was processed with Sequest and either Peptide Validator or Percolator using the manufacturer's default settings.  Each processing scheme was ran twice (alternating between schemes).

Output
Peptide validator: 
Processing time:  7 minutes
Peptides:  216
Peptides, unique: 201
Protein groups:  61
Proteins total:  75

Percolator
Processing time:  18 minutes
Peptides: 401
Peptides, unique: 381
Protein groups: 91
Proteins total: 120

I exported the peptide report to Excel:  Percolator found EVERY peptide that Peptide Validator did, and simply added 179 new peptides.

The next question, of course, is:  are these new peptides any good?

DigDB

Due to some changes at my work, we are no longer allowed to make purchases using PayPal.  Unfortunately, this also corresponded to my DigDB license expiring, as well as the elimination of the free Google Checkout option.
Luckily, the people at DigDB are pretty nice and good to their loyal customers.  They gave me a 6 month subscription to tide me over until we figure out a way to pay them.
If you are doing any kind of metabolomics, genomics, or proteomics work and you don't have DigDB, I sincerely recommend you spend the $78 for a 1 year subscription or at least give the 15 day free trial a whirl.  This add-in gives you a ton of new functions in Excel including the ability to combine entries and number the occurrences.
For example, if you are looking at a complex list of sequenced peptides or phospho-peptides, you can combine the Thermo output spreadsheet so that you get all of the variations of a single peptide all in one easy row.  It will add a new column that tells you how many incidents you have for that peptide (it also orders them from the most commonly occurring peptide first).  You then can expand that row to see the variations of that specific peptide you obtained.

You can try the software out here. They also have some really easy and simple tutorials.

Orbitrap transient (scan) times

This is a chart I made using various pieces of literature out there on the various Orbitrap systems.  I don't know how accurate all of this is, but this is my understanding.  Keep in mind that scan times do not include fill times.  My Orbi Velos can do a 15k scan in 192 ms, but if it took 200ms to collect the requisite number of ions for the scan, I'm only going to get a little more than 2 scans/second.

Today's Lunch Read: Quantitative Proteomics Reveals New Insights into Erythrocyte Invasion in Plasmodium Falciparum

Today's lunch time fun was an article that came out in this week's Molecular and Cellular Proteomics.  The group is from Nanyang Tech in Singapore and you can find the abstract here.

The title of my post describes the paper pretty well.  They used a combination of transcriptional and quantitative proteomics (iTRAQ) to interrogate some P.f. cultures that were treated with neuraminidase.

The iTRAQ proteomics samples were analyzed with a QStar Elite.

The highlight of the paper is the two very nice tables of the up- and down- regulated proteins found by iTRAQ (roughly 75-100) when the cells are treated with this agent.

While this paper is a nice solid study, it does a good job of highlighting how little we really know about this organism, as well as how relatively poorly the genes/proteins are currently annotated.  

Percolator in Proteome Discoverer 1.3

A new feature in Proteome Discoverer 1.3 is the Percolator program that was previously bundled with the Sieve program.  I have been running a few data sets with both the Percolator and with Thermo's Peptide Validator to determine which would be the best for our datasets.  One run of 48, reasonably complex samples, that I performed over the weekend ran like this:
1) Started processing around 4:30PM (sending results to the Mascot at our local cluster)
2) Mascot results received a little after 9:30PM (5 hours)
3) Percolator ran for 8-12 HOURS!  I am uncertain about the time since PD claims that Percolator ran 8 hours 41 minutes, but it started around 9:30 PM, but the clock says it finished around 9:00 AM.  Weird 4 hour time differential I can't possibly justify today.
4) Itraq reporter ions quantified and assigned to ions in roughly 20 minutes.

This is performed on a Power PC with a Xeon Quad-Core with >50 GB of RAM and striped hard drives.  Optimized for pure computational speed.  I'm currently re-running this same data set using the Peptide Validator.  If the data is not significantly better with Percolator, there is no way I can justify this additional time for processing.

Researching Percolator has identified this paper as the original reference.  I'll post more after I work through it (probably lunch time reading today).

Current reading: Post-translational modifications in Plasmodium: More than you think!

Today's lunch time reading (performed at a nice Korean restaurant) was a Review in Molecular and Biochemical Parasitology.  The PubMed abstract can be found here.
This paper is an excellent review of what we currently know about the regulation of Plasmodium species.  Unsurprisingly, from what I have learned about this organism so far, there seem to be more questions than answers.  However, we know that these parasites can regulate themselves by phosphorylation, ubiquitination, acetylation, methylation and lipidation, just like other eukaryotes.
The differences highlighted by this paper are in how these processes are regulated.  For example, the parasites lack MAP kinases and tyrosine kinases similar to other eukaryotic ones.  These kinases are so conserved in every organism I've ever looked at that I assumed they were the same in all species.  Despite the apparent absence of these incredibly important kinases, phosphorylation is still very important to Plasmodium since their growth cycles can be halted by treating them with certain kinase inhibitors.  It will be interesting to see what kinases are important, and if these kinases are present in the host organisms, as parasite-specific kinases would make very interesting drug targets!

Proteome Discoverer 1.0, 1.2, and 1.3 comparisons

I finally got to sit down and run multiple datasets side-by-side on Proteome Discoverer 1.0, 1.2, and the new 1.3.
The results are pretty much what you would expect, but let me describe the data set:
-One run of serum taken from a mouse, depleted of the 4 most prevalent proteins (albumin, IGg, transferrin, and one other), digested, separated on a 240 minute gradient and ran using my standard Top10 CID method.
-All modifications were the same
-All cutoffs were the same
-The data was sent to both our Mascot server and the local Sequest program
-The default FDR was employed on both 1.2 and 1.3 (no FDR in PD 1.0)
Results:
PD 1.0:  408 protein IDs
PD 1.2:  121 protein IDs
PD 1.3:  275 protein IDs.

Wow, right?  I'm still investigating, but my hypothesis is:  PD 1.0 has a tremendous number of false positives.  Realizing this, Thermo added FDR calculations to PD 1.2.  These were so strict that they cut the data received dramatically.  PD 1.3 employs a new algorithm that is closer to the true peptide counts.

I will continue to investigate as I have opportunity.

Orbitrap Velos ETD vs. Orbitrap XL ETD

So, I've been using the new Orbitrap Velos plus ETD for about 2 months now and these are my observations in comparison to my trusty old Orbitrap XL plus ETD.
1) The shorter ion transfer tubes on the Velos get fouled about 4 times as fast as the longer ones on the XL
2) The file sizes on the Velos are approximately 1.5 times larger for the same sample as the same sample ran on the XL
3) While the number of proteins ID'ed on the Velos don't seem to exceed those of the XL, the individual peptide scores (and thus the total protein scores) are higher when ran on Proteome Discoverer 1.2
4) Although the XL plus ETD is supposed to have the same HCD cell as the Velos plus ETD, I have experienced dramatically better HCD response.  For example, I always found that the XL wasn't all that great with reporter ions such as iTRAQ and TMT, where these technologies work extraordinarily well on the Velos.
Hopefully I'll be able to run some samples side-by-side on the two machines in the near future and see how the coverage and peptide scores differ!