Georg Wallmann

We just release alphaDIA 1.8.2 including some new step-by-step guides for #Proteomics. Learn how to: • Run DIA searches from the GUI • Create custom PTM models with DIA transfer learning • Evaluate model performance with detailed metrics alphadia.readthedocs.io/en/latest/guid… #TeamMasSpec

@wfondrie @nesvilab @i000 @bruker @jspaezp1 To my knowledge we are still lacking some additional proprietary calibration with timsrust/alphatims on Unix when Bruker DLLs are not available. What’s your experience?

@nesvilab @i000 Also, kudos to @bruker for supporting open-source with timsrust: github.com/MannLabs/timsr… as @jspaezp1 mentioned. (Disclaimer: Talus does have a collab research agreement with Bruker and we do contribute to its development)

With some recent discourse in the #proteomics community around #OpenSource scientific software, I thought it was finally time to jot down my thoughts. I wrote a quick blog post, complete with my own 🌶️ takes. Here's some of what it contains: (1/2)

@DemichevLab Ah okay, I see 👍🏻thanks

@GeorgWa Hi Georg, this is described in the original 2020 paper, it’s the ‘no shared spectra’ checkbox

Please see our HUPO poster on achieving peptidoform confidence in DIA proteomics, now online github.com/vdemichev/DiaN…

@ucdmrt @michaellazear @fcyucn @animesh1977 @chrashwood @maxquant @github @JimmyEng Yes sure! Best is if you reach out on GitHub discussions: github.com/MannLabs/alpha…

@GeorgWa @michaellazear @fcyucn @animesh1977 @chrashwood @maxquant @github @JimmyEng Thanks Georg, I haven't tried it yet but I will give it a shot with some Exploris samples and can let you know how it goes ?

@ucdmrt @michaellazear @fcyucn @animesh1977 @chrashwood @maxquant @github @JimmyEng I can only speak for alphaDIA but we are actively developing it and get very competetive performance for all Thermo or Sciex instruments as well as library based TimsTOF searches. Feel free to reach out if you want to discuss some use case.

@michaellazear @fcyucn @animesh1977 @chrashwood @GeorgWa @maxquant @github @JimmyEng Experience with on Alphapep / AlphaDIA from Mann lab?

@jalew188 @Westlake_Uni @labs_mann Well deserved! It was a great pleasure working with you and I'm looking forward to future collaborations 😀.

I have officially joined @Westlake_Uni as a tenure-track assistant professor. I will be returning to the pGlyco project to make it more accessible, not only for the software but also for the source code. Thank you @labs_mann for these memorable four years, and for the nice gift.

@ItsBiniR No need for libraries. This is a perfect application for end-to-end transfer learning. It allows you to train your own instrument and modification specific model right from your DIA data: biorxiv.org/content/10.110…

Cysteinome profiling by DIA using a custom spectral library of DBIA-bound peptides generated from 72 mass spec raw files.. Data from HEK293T and Jurkat cells were used here.. Can this library be used for unrelated cell lines?? If not, how practical is DIA for chemoproteomics?? 🤔

@chrashwood I know another feature free DIA tool ;). Sounds like this is hot at the moment.

Cellular states prior to stimulation will be predictively associated to outcomes. By linking variable proteomes to responses, we may identify regulators. biorxiv.org/content/10.110… Here, we use single-nucleus proteomics to identify regulators of LPS-induced protein transport🧵

@Hamish_Stewart @mjmaccoss @DemichevLab Isn't this already somewhat accessible by looking at SNR ratios and fill times?

@mjmaccoss @DemichevLab Agreed. When one really has number of ions detected, mass positions etc, then it can't be difficult to show drifts. Maybe this is an argument to save and expose the Astral current measurement scans, though we were worried about confusing bioinformatics software.

A note about routine LC-MS performance quality control. If you are doing DIA, then any decrease in MS sensitivity is best detectable with low sample amount injections (less than the reasonable maximum amount by a factor of 10 or more).

A novel hybrid high speed mass spectrometer allows rapid translation from biomarker candidates to targeted clinical tests using 15N ... biorxiv.org/cgi/content/sh… #biorxiv_biochem

1/7 @labs_mann and Georg Wallmann @GeorgWa are proud to announce alphaDIA, our open source, next generation DIA search engine. Check out the preprint @biorxivpreprint biorxiv.org/content/10.110…

AlphaDIA enables End-to-End Transfer Learning for Feature-Free Proteomics biorxiv.org/cgi/content/sh… #biorxiv_bioinfo

@OliverMBernhar1 @labs_mann I think Spectronaut should qualify you 🤔

@labs_mann Does "I build it from scratch" imply one wrote a DIA search engine? 🤔

We've got something exciting to share before #ASMS2024! But first, how confident do you feel about your DIA search? #TeamMassSpec

@MartinFrejno @msaid_de What external benchmarking would you like to see to feel confident about an identification from MS1 only?

@GeorgWa @msaid_de I would be very cautious with calling an identification based solely on AMT Tags, even on single cells. We are not doing AMT Tags anymore for a reason. And with shorter and shorter gradients, you substantially reduce the information content of the RT dimension.

Hello #TeamProteomics and #TeamMassSpec! We're gearing up for #ASMS2024 with some intriguing questions on #Proteomics and #FDRControl. What is the minimum number of fragments you consider needed for the identification of a precursor? Join the conversation and let’s dive into this together! 👇

@mjmaccoss @MartinFrejno @msaid_de @SciFiRobitaille I very much agree with this notion. I think confident identifications with few or even none fragments are well possible if there is enough evidence. The real challenge is quantification and giving good estimates of quant certainty.

@MartinFrejno @msaid_de I mean, it's just idea :D. I'm thinking of something like an AMT Tag. Just not with a proteome specific library but a whole proteome library. You wouldn't get the same performance in complex bulk proteomes, but maybe it's something for single cell.

@GeorgWa @msaid_de Now that's a statement! But why would you then need a library in the first place? Or are you implying performing identification solely based on the MS1 signal (i.e. AMT Tags)?