Why are my MSAmanda 2.0 searches faster?

I recently mentioned in a Twit on Tweeter that I've just almost moved to MSAmanda exclusively for my data processing. The one final complaint (for me) is now gone. The newest version is so fast that SeQuest is irrelevant.

Multiple great people responded that they aren't seeing the same thing. I figured it couldn't hurt to get my settings and stuff out there so we can try to figure out the difference?

Probably worth mentioning that I'm searching a highly fractionated data set ran on an HF-X (48 files and around 2.7e6 MS/MS spectra for dynamic mods of K+acetyl, S/T + phospho and M+ oxidation. Even with ptmRS and Percolator the total search time is something like 4.5 hours.

It's this version:

Important -- there is a big difference in speed when using second search (you're removing all the fragments that match your first peptide and then going back through and re-searching every MS/MS spectra for new peptides. Go over to "Show Advanced" and scroll to the bottom. Make sure it looks like this if you're doing a speed comparison. If you want to identify more peptides, particularly if you're doing single shot -- switch it to True!

For this search I'm using 5ppm MS1 and 15ppm MS/MS (yo -- don't tell anyone, but I didn't run this HF-X at full speed. I needed more fill time anyway -- interestingly there is a huge difference in the number of lysine acetylation sites I recover if I run the same sample at transient match vs excessive fill times. [30ms vs 150ms]. Its in a paper that I have been spending every waking moment addressing reviewer comments on and the main reason I haven't blogged in a like a month.... oh -- and I honestly don't like the HF-X mass accuracy at full speed...Maybe it's this particular instrument? However, it's similar to my experiences running a Fusion 1 at 8,000 resolution, I can't reliably use 15ppm MS/MS fragment accuracy. Again, could be just this instrument. Could be a calibration issue, but at MS1 5ppm on this dataset I have no problem, either way I went with 15,000 resolution MS/MS for this study)

And that's about it. If I run Sequest the same way (but with 0.02 Da fragment) MSAmanda is faster.

If the PC specs matter?

I'm using an OmicsPCs Maximum Destroyer I purchased at ASMS Indianapolis. Unethical life pro tip (ULPT): If you're local to a conference and drive a hatchback to it, sometimes you can get deals by going around to the booths and asking people if they really want to ship that used piece of equipment home. 10 minutes before a show closed up this year, we got 2 of these things!  How do you homogenize a solid tumor? There are lots of ways, but you could start by putting it in an Omni Beadruptor12 for 7 (SEVEN) seconds. Full disclaimer: I haven't had time to compare the extraction efficiency for proteins, but we will soon. Small molecules? You can extract them out of wood in a minute or so. Not kidding. Like the wood from a tree or that you make furniture out of.

This PC is a couple of years old now, but it still maxes out Windows 7's bizarre (1-7.9?) rating scale. It appears to have 20 threads. I assume it's 10 cores and 10 simulated "hyper cores/threads" or something, but I don't know.

I have put in 2 upgrades myself. It has a 1 TB Samsung Evo (I think it originally had a smaller Samsung Pro Drive) solid state drive and I put a 1080T graphics card in it so that I could explore a simulated universe that 1.8e19 planets in it at a high frame rate while waiting for MSAmanda 2.0 to finish searching stuff.

The GPU does nothing in Proteome Discoverer, solid state drives are critical for fast searching. I don't know if the RAM on this thing is a special high speed configuration or something. I can see through the glass that there are 4 sticks and they pulse blue LEDs off and on, and they do it much faster when a search engine is running. (The speed increases almost up to strobing when I run MetaMorpheus searches, does that mean MM uses tons of RAM? I dunno, but that would make sense). 

I assume none of this was helpful, but I sure missed typing in this box!

MeroX 2.0 -- It's ready for big time global stuff now!

In the "return"(?) of MS crosslinking (? cause, did it ever leave? probably not, but there was definitely a point where we all avoided it whenever possible) a familiar name kind of faded into the background. StavRoX (non MS-cleavable) and MeroX (MS-cleavable) also never went away, but other newer software gained a lot of ground with new concepts. This new study shows that the software team didn't let these new improvements pass by un noticed

Earlier this year(? I lose track of time easily) Rappsilber lab demonstrated that monoisotopic precursor selection (MIPS or "Peptide match") doesn't work very well for crosslinked peptides and they adjusted for it in their software.

MeroX now does the same.

A couple of studies have shown that XlinkX can really handle global level eukaryote crosslinking data. Improvements in MeroX 2.0 can now search entire organisms as well! (They demonstrate by finding over 1,000 crosslinks in fruit flies)

I'll be honest, I don't entirely entirely understand false discovery rate (FDR) applications in crosslinking and how/why these differ and why the traditional stuff doesn't work right, but it appears that Sinz lab has developed a crosslink specific FDR. With so many published software packages fading into obscurity, it's always great to see one refreshed by a team paying attention to all the ways they can make their software better. It's all available at StaVroX.com

Ribo-Seq + TMT Peptidomics!

Just this summer I heard about Ribo-Seq (WikiPedia article here) for the first time.  Honestly, Ribo-Seq or Ribosome profiling is one of those technologies that I think will start taking lots of the easier work from proteomics facilities -- if it hasn't done so already. You get a ton closer to accurately predicting the protein levels if you're actually measuring what gets to the ribosome, right? 

In this new study the two are used in unison to look for new brain specific neuropeptides.  

The approach used here seems to combine the results of TMT proteomics and RiboSeq and look for concordance. It is worth noting that another approach (Proteoformer 2.0 post here) that is a bit newer than this March publication has demonstrated combining the two a bit further upstream to amplify the results.

In the Nucifora Lab study we see that the two are highly complementary and have the ability to differentiate organ specific peptides and proteins.

Don't throw out all the Keratins! They might be important!

Now that I'm back in the part of the world where you can't see stars at night, but you do have internet service and for no particular reason, I might highlight research done right here in rat infested Baltimore.

This is a short one because I've got to get my fingers used to typing again, but this cool recent paper in Science Translational Medicine highlights why you shouldn't use a filter that throws out the word "keratin" in your results! 

The patients that have this disease have massively more of some weird keratins....

For real -- I don't think I could have caught this with several of my default filter setups. Particularly if I was doing skin proteomics? I might have tossed every word that started with a K....

ABRF GenomeWeb Follow up.

If you missed the ABRF GenomeWeb talks we did the other day, you can still watch them on demand by going to this link and registering. Since my animations didn't work, you can get my slides with working animations here.

We went a little long and didn't get to all the questions. I took screenshots of them and have been working on them.

Q1: Is there something like RAWMeat that can monitor specifics of instrument performance longitudinally?

A1: Yes. There are several, but these 2 are my favorites:

RawBeans

(New!) QCloud

AutoQC and sPRoCoP have to be mentioned as well. There's loads on this dumb blog about them.

Q2: TMT vs DIA comparisons with today's instrumentation? How do they compare?

A2: It sounds like Dr. Schilling's group is doing a massive and comprehensive comparison and we should watch out for it, but this is a great recent paper where time constraints were utilized as a parameter:

Q3: Can you use a Q Exactive Classic for DIA? 

A3: Totally! Is it as fast as the high field stuff? No, but once you adjust for these limitations you're set. Here is a great example:

Q4: Is there a reason to generate libraries now that ProSit is out there?

Uninformed Opinion 4: Scott Peterson did a really extensive analysis years ago, I have slides here somewhere and ALWAYS in house generated spectral libraries were better than any in silico models that he tried. It wasn't even close. There is individual instrument and lab and sample variations. ProSit is a big step forward. These new intelligent machine based libraries are better than anything we've ever seen, but I bet you that in house generated will still be better. I don't have data to back this up. I hope that the margin is small. I'll definitely go to Prosit first. Instrument time isn't free. However, I'd comfortably bet $4.13 that when the dust settles and we see the inevitable 10-15 papers that compare the two in the next 18 months or so that in house is at least marginally better. (I do hope I'm wrong, though)

Q5: How much longer do you think that MS will be the dominant technology in proteomics? 

Uninformed Rambling Opinion 5: It depends on how we define proteomics. Are we talking peptide abundance? If so, then I don't expect MS to be the leader for much longer. Arrays are coming. Nanopores are coming. If we're defining proteomics as modified protein abundance and/or top-down, intact protein analysis and quantification? I think we can safely say that MS will be the dominant force for the remainder of my career.

Arrays are neat and everything. And cheap. And they'll get better, but you have to know so much in advance to use them. I'd guess only a few more years before it is better/cheaper/faster to use arrays if you want to find out protein abundance. Honestly, though, who cares about protein abundance? Really? RiboSeq is getting a lot more accurate and faster/cheaper.

We might even see arrays replace phosphoproteomics. I'd have a party. Let's give phosphoproteomics to someone else, because, at the end of the day we all know that LC-MS/MS isn't very good at it. Not really.

Where LC-MS/MS has a tremendous advantage is all the other PTMs that we've only barely explored. Glycosylations, acetylations, succinylations, Sumoylation, Ubiquitinations and our ability to look at all of them. And intact protein analysis on a global scale? Yeah -- it's getting closer all the time and no one on earth wants that problem except for us.

WinProphet -- Trans Proteomic Pipeline for Everyone!

I'd like to offer a big thank you to these authors! The Trans Proteomic Pipeline is something I've always respected and read lots about, but have always been far too dumb to use for myself.

You know what I need? A friendly Graphical User Interface (GUI) and that's what WinProphet is.

BOOM -- Access to all sorts of tools you've probably heard about and you were probably smart enough to use it (I'm not) but you didn't have time to learn a new command line program thing with all the options you have for searching proteomics data.

You can skip all the reading and just get WinProphet here!

Expert Tips? Running a Mass Spec Core Amid Rapid Change Webinar Tomorrow! 1pm EST.

If I did this right again this time, you can register for this ABRF/GenomeWeb Joint Webinar by Clicking anything from this line up.

And....so...it's obviously totally July right now. Not only July but also like past the middle of July and I'm 100% prepared for things that are to happen in the middle of July like my meeting with some super important government persons today about the second lab I've built from scratch in less than 5 months, as well as giving a webinar for freaking GenomeWeb with 2 of my personal professional heroes and of course my webinar is totally done and suitably well rehearsed....

....you should definitely register, the other 2 talks are great and, at the very least, I'll try to say something funny, but I'll try to also come up with something smart. (This is a recap of our ABRF 2019 workshop with updated material. I definitely know where those slides are, but it enjoyed writing it this way more.)

Biden Cancer Moonshot suspends operations....

I hope this doesn't have consequences for people funded by this amazing program, but conflict of interest used to be a real thing our government considered.....

AP story is here.

Ben's inaccurate guide to FDR in Proteome Discoverer -- what to use and when!

Every once in a while I start a blog post and I think "oh no. what am I doing?" This is probably going to be one of them, but I'm probably going to keep typing, but keep this in mind.

1) I have well over my 10,000 hours of kicking data of all types around with different search engines and false discovery rate (FDR) algorithms -- so I have real world advice.

2) This involves statistics. My last formal course on statistics was in the 90s. I did tutor statistics. Also in the 90s. We didn't use computers in statistics in my undergrad institution -- because it was the 90s...

Inaccurate FDR guide time! 

Lets' use the Proteome Discoverer stuff above because I actually know about it most. Let's start with the easiest one, using PD 2.2's fantastic and logical numbering system

4) Fixed Value PSM Validator = Translation: No FDR!

What is it? That's actually inaccurate in the later versions, but up to PD 2.1 (I'm pretty sure) there were no settings in this box. It is basically there because you have to have something. This relies on the cutoffs within the search engine itself. If SeQuest, you've just moved the scoring to entirely rely on the XCorr cutoffs within your Admin --> Configuration --> SeQuest --> Node

In PD 2.2 and 2.3(?)  this now includes your DeltaCN cutoff. DeltaCN is just detailing what you do when you have 2 sequences that match your MS/MS spectra. Imagine you have these 2 matches and one gets a score of (bear with me) 100 and the other gets a score of 96. If your DeltaCN is 0.05, you get a peptide spectral match (PSM) for both sequences, because your delta off of your highest match is within 5% (0.05). However, if you get 2 potential peptide matches and one is 100 and the second is 84, then only the top PSM is reported.

When do I use it? When I'm looking at a single protein, or 5 proteins, or 10 proteins. Maybe up to 100 proteins? Not many proteins! Or if I want to get a ton of IDs from a global dataset and I don't care about the level of accuracy in those IDs....which is a joke. I don't use this for global and you shouldn't unless you've got other filters like >3 unique peptides per protein or something.

6) Target Decoy PSM Validator = Translation: The first kind of FDR for proteomics

What is it? It was a revelation in proteomics about a decade ago. Unless I'm mistaken it was first described by Elias and Gygi? Although, the paper I typically associate it with is this paper from the same authors a few years later.

You essentially do 2 searches:
1) Normal FASTA database
2) Your FASTA database that you screwed up. You either read all the entries in reverse or you scrambled them.
THEN
3) You adjust the scoring on the two searches to cutoffs that only allow a small percentage of the screwed up database to match. For example (and yes -- I know -- this is only a useful statement for visualization...I've heard all the statistician tirades...) you adjust the scoring until only 1% of your screwed up FASTA database gets picked up as real hits. This is the whole misnomer of 1% FDR. Never say this in front of an audience.

The assumption you lean on is that your screwed up database won't get matches, or will get very very few. On a big database that has been screwed up, there are ALWAYS going to be matches by your search engine. The goal here is to control the random chance of PSMs occurring (this works if you assume the scrambled database matches would only occur by random chance, which - due to things like organisms biasing toward specific amino acid usage and the biological value of having repeating sequences in some proteins -- isn't really true, but it is uncommon enough that it makes a good metric).

When do I use it? When I need to be conservative with my identifications. Particularly for big databases, target decoy is going to throw out a LOT of good data. I do not typically use this for single proteins. Maybe I do as a comparison?

(Too much uninterrupted text in a row makes me nervous. Pug Pile!!) 

5) Percolator = Translation: Try to find the good data that target decoy threw out by looking at a bunch of factors.

What is it? It is essentially the first foray of mainstream proteomics into intelligent learning machine deepness (insert arbitrary sounding big data computer term here -- again -- check WikiPedia a LOT before you use any of these terms for an audience. They all sound like the same thing but they aren't. Percolator is really a "semi-supervised learning" thing, and people take these different terms very seriously. I once made the innocent mistake of referring to some Harley Davidsons (sp?) outside a bar as "nice bicycles" inside said bar, which, by the way, they definitely technically are and it was almost as poor of an interaction over a technicality as I've seen at conferences where people got their artificial learning things mixed up.)  Percolator was probably first described here.  A while back a colleague assembled a list of all the things that Percolator considers when rescoring a Peptide Spectral Match (PSM) and that list is here.  Since Percolator rescores peptides with it's metrics its great for combining the results of different engines.

When do I use it? When I've got a LOT of data and I want to get the most identifications possible. I NEVER use Percolator for a single protein digest. I don't know the realistic cutoff where I trust Percolator output. I might put it at around 1,000 proteins. BY THE WAY. If someone gives you an immunoprecipitation or Co-IP and it's got 4,000 proteins in it, there are lots of great guides online on how to do one of these experiments properly. Ask that person to read one and try again. Don't spend all your time trying to interpret garbage pulldowns. If you are going to try and make sense of it, don't treat it like a Co-IP experiment. Treat it as an affinity enrichment and just do everything they do in this paper.

7) IMP-Elutator = Translation: Percolator + Chromatography data

What is it? As best as I can figure out, Elutator takes the Percolator approach but then also incorporates chromatography information predictions. I think IMP-Elutator was first detailed here.

When do I use it? Big datasets with a lot of PTMs. MSAmanda uses more sophisticated features for PTM identification and confidence in those identifications than older code like SeQuest. MSAmanda rewards your scores for higher mass accuracy (measured in PPM accuracy, rather than mass range bins) and can handle things like neutral losses from your peptides (akin to MaxQuant searches). Since I'm already using MSAmanda (2.0!) for getting high confidence PTMs, I go ahead and use the tools that were designed to work with it. More importantly, as you've added PTMs, your search matrix increases exponentially, as does your likelihood of getting bad matches. By taking chromatography into account you can further improve on your FDR.

Okay -- I hope this helps? Thanks to Dr. R.R. who wrote me the question that made me realize I hadn't updated anything on FDR in PD since like 2012 and I should reeeeeaaaallly go back and delete some old stuff!

InterPro and UniFire -- UniProt is more than a place to download FASTAs!

Sorry for all the "building FASTAs and genomes from proteomics and nextgen stuff" posts. If your organism is all annotated and clonal this is probably boring." However, if you are studying something that hasn't been perfectly annotated or has individual genetic variation this is a great time for you! So many options!

InterPro isn't new, by any means, but InterProScan (scan is the web portal) was just upgraded last week and EMBL-EBI has some very convincing arguments (like near the end of this YouTube video) for why you should be using it for your annotations.

Okay -- having a gorgeous web interface shouldn't be the reason to use something -- but when it's backed with this kind of ridiculous power, it doesn't hurt. All these databases feed in seamlessly and you can choose to use some or all of them.

Like the pBLAST web interface you can only search one thing at a time, which is great if you're down to a small list of quantitatively interesting things. However, there is a stand-alone program (InterPro, I think). It can only be installed in 64-bit Linux, so no casual Windows people.

Okay -- but the reason I've been running InterProScan is actually because of something else from EMBL -- it's called UniFire and there isn't tons out on it yet.

However, there IS a free class on Wednesday July 10th on UniFire. You can register for it here.