I was wondering why those volcano plots are weird. Was it something related to Olink? But then I saw the methods and remembered my criticism of the vibe coded algorithm in JPR using the z-score. Bad statistical decision in Nature Medicine as well. 🫣
70% is a fantastic improvement. Thanks for the algorithm.
It wasn't. There is a pull request for this from Oct 2025. Here:
share.google/YNx5yWUm9nDH...
Still not there. The column is null in the example and is null in my output from the updated version.
Great tool, I'm still waiting for the add of the retention time values in the mzTab output.
Yes, context is important.
In this case, the customers are scientists interested in solving specific problems. And that's how things progress, with people solving problems.
Good user interfaces, good integrated post-processing tools, etc. This will bring advances in the de novo search proteomics field. Because people will use it.
Let's be honest, creating an algorithm that requires console operation and delivers a report with gigantic, dry tabular data isn't a good way to achieve popularity. The maturity of database search algorithms came with this understanding.
Algorithms don't exist in a vacuum. People use them. An algorithm should be easy, flexible, and intuitive to use. Think about that. You can argue that researchers should know how to use Python, C++, Rust or whatever you run in your lab. The truth is that most of the PIs don't know how to do it.
A harsh truth about developers in the de novo proteomics field: they don't care about user experience. Do you know why Windows (~60%) and macOS (~15%) are more popular than Linux (3%) in the desktop computer world? Because they care about user experience!
To be honest, I don't feel like de novo developers care about it. There is no minimal effort to make de novo popular.
Most algorithms are quite incomplete: casanovo does not report RT and intensity, and although it can do protein inference, it is extremely slow. Instanovo does not report intensity or protein inference. Proprietary software like Peaks is extremely expensive.
You can do nice things, such as this one. I will have something really cool to share soon.
Not actually, the data goes to the psm.tsv. a couple of new columns and fragments concatenated on the single cell. Perfect. I can use the same function as I do for MQ.
found it! The checkbox "Export matched fragments" in Run tab. Why is it not checked by default? It is exactly the same information.
Do you mean that interact-sample.pep.xml file? If so, it is not as useful as the msms.txt from MQ. No structure, not complete. I guess there is no equivalent file, unfortunately.
Is there a file generated by FragPipe that stores information similar to the msms.txt file from MaxQuant? That is, matches (b and y ions), intensities, and m/z values. If not, adding this information would be invaluable.
#FragPipe
ending the bypass of tsv file, quality-of-life
Intrusive thought of the day: a proteomics publication using DIA-BERT to investigate a cohort of individuals with diabetes.
www.nature.com/articles/s41...
it is interesting to have an idea of what can be discovered with de novo in terms of modification. I will read your work and maybe try both approaches. I'm thinking on a kind of 2-steps search: first open, then constrained.
People doing de novo proteomics. Did anyone test this idea of open search for de novo search? I'm planning to play with it and report some toy dataset here in the next few days.
github.com/pFindStudio/...
I'm waiting for this integration with InstaNovo+
A nature-inspired ion trap for parallel manipulation of ions on a massive scale www.science.org/doi/...
---
#proteomics #prot-paper
QC4DIANN allow you the load the report.parquet from DIANN and export the protein abundance matrix filtered by the QuantUMS scores and use it in Perseus or only other pipeline. Added the option to save all the figures in 300 DPI.
BTW: You may have noticed how much I like DIANN and FragPipe :)
DIANN users: making your life easier when evaluating your DIANN search results. QC4DIANN_v1.1.0 has new visualizations and now accepts the fasta you used to generate the speclib as a reference for calculating the proportion of proteotypic peptides per sample. Release 👉 github.com/41ison/QC4DI...
FragPipe users: I've updated PSManalyst shiny app. Check it out and enjoy the easiest way to evaluate the quality of your runs. Now you can map peptides to protein coverage counting PSMs.
github.com/41ison/PSMan...
I don't think so. In this case, I would do pure DDA for non-tryptic peptides
I would run a DDA with de novo search in casanovo with non-tryptic ckpt model.
