A Ground-truth validation of FDR & false localisation control in proteomics – a must-read from Stefan Tenzer’s lab!
The benchmarks are quite enlightening. We are pleased to see our DIA-NN 2.0 excelling in sensitivity - often by a wide margin - while controlling FDR and false localisation rates.
Solving the computational challenge of phosphoproteomics with 𝐏𝐡𝐨-𝐓𝐢𝐩: One-Pot Dephosphorylation for Rapid and Sensitive Analysis of DIA Phosphoproteomics Data. Now out in Analytical Chemistry!
Makes predicted phosphopeptide libraries 10x-20x smaller. Link below.
Naive question: I thought targeted was about sensitivity -
why ng?
It's called the Agilent 6495D.
Not a conference, but worth flagging:
EMBO Practical Course “Targeted proteomics: advanced tools for biomedical research” Barcelona, 8–13 Nov 2026
Line-up not announced yet, but previous editions organized by @maccoss.bsky.social with invited speakers incl. Alexey Nesvizhskii and Vadim Demichev.
Proteomic Ruler question:
In Wiśniewski et al., MCP 2014, the histone→DNA proxy seems implicit.
Is there any explicit reference stating that the Ruler uses only core histones (H2A/H2B/H3/H4) and excludes H1?
#proteomics #massspec
I agree MaxLFQ isn’t meant for absolute quantification.
But that still doesn’t explain the complete lack of correlation with UPS2.
From my experience, iBAQ and MaxLFQ usually correlate well (R² ~0.78, non related example dataset shown), suggesting they track the same MS1 signal.
As shown in the DIA-NN paper, the mobility term contributes only negligibly to the discriminant score, suggesting that measured CCS -even with good IM resolution - might simply be too affected by gas-phase ion–ion / ion–neutral interactions to provide a stable, high-specificity constraint …
With a quadrupole we know exactly which precursor m/z window was isolated -the precursor mass is tightly defined with a well-characterised error. My question was whether CCS can provide anything close to that level of search-space restriction for database searching.
My point rather was whether precursor CCS can actually constrain the search space during database searching not just what Da-equivalent tolerance it has.
Asking for a friend: Is intrinsic specificity of CCS high enough to serve as an effective in silico precursor filter during database searching?
“Very interesting! Do you know whether ProteomeSciences is already testing the new DXT tags with selected customers or collaborators, or is it still entirely in-house at this stage?”
Bonus, info about DIA multiplex tags, up to 30-plex:
"trademark DXT for our DIA multiplex tags...advances have been made in DXT multiplexing since ASMS with the number of tags increased from 6 to 11 and with the potential to
increase these to beyond 30"
Fantastic project led by @bo-wen.bsky.social. Excited to see the future uses of AI and transfer learning in proteomics. #massspec #proteomics
www.nature.com/articles/s41...
Which LC & Flow?
Surprised that u go so low. With EvoSep 24 min method we can load lots more on our Ultra2 until reaching saturation especially with ICC2.0
How much are you loading per injection? Is ICC 2.0 enabled on the Ultra2? And which library are you using?
True — but the odd part is that the Human Reference Proteome is not really ‘canonical only’. Non-canonical entries from TrEMBL are included, yet the curated SwissProt isoforms are missing default. That’s what undermines the idea of a high-quality reference set.
On a separate note: I was surprised to find that none of the non-canonical SwissProt isoforms are included in the official human reference proteome (UP000005640).
Anyone know what’s going on here? 🤔
#proteomics #bioinformatics @pwilmarth.bsky.social il
By the way.. 43% of current TREMBL entries will be dropped soon anyway insideuniprot.blogspot.com/2025/06/capt...
Hey #TeamMassSpec,
Many non-human proteomics studies still search against taxon-filtered FASTAs.
❌ Redundant sequences
❌ Inflated search space
✅ Reference proteomes cut redundancy, improve annotation, and make results comparable.
👉 Time to move beyond taxon filters. #proteomics #massspec #uniprot
Without #2, a lower ion count is needed just to be sure that the full MS range is scanned, but with more accurate ion counts, you can go to the max S/N without losing ions on the edges.
This could also work for the Orbitrap Astral.
Bonus: DIAPASEF on Thermo - patentscope.wipo.int/search/en/de...
With 𝗗𝗜𝗔-𝗡𝗡 𝟮.𝟯.𝟬 Preview (Academia-only for now), we showcase the transformative new capabilities that have been developed in the past months. Download: github.com/vdemichev/Di...
Thanks @pwilmarth.bsky.social - also included the less-redundant "one protein per gene" db here ...Has anybody assessed potential benefits of the reduced search space on sensitivity?
aaah guess its "hidden" there :)
Great thanks! Where can I find the one protein per gene option?
Hey #TeamMassSpec,
When you run proteomics on non-human species (mouse, rat, macaque, etc.) — which protein FASTA do you prefer?
Taxonomy-filtered UniProt (all entries)
Reference proteome (SwissProt+TrEMBL)
Ensembl/GENCODE
Something else?
💯
Astral Zoom hits >7,000 protein groups & 67,000 precursors — on a 500 SPD EvoSep ENO run.
www.biorxiv.org/content/10.1...
Is that narrow windowed thin-PASEF?
