Alison Chaves
@alisonchaves.bsky.social
170 followers 220 following 240 posts
Doing proteomics until I still can. https://github.com/41ison
Posts Media Videos Starter Packs
Pinned
alisonchaves.bsky.social
It's now properly published. If you want to easily check important characteristics of your data before diving into complicated statistics, check out PSManalyst.

PSManalyst: A Dashboard for Visual Quality Control of FragPipe Results | Journal of Proteome Research pubs.acs.org/doi/10.1021/...
PSManalyst: A Dashboard for Visual Quality Control of FragPipe Results
FragPipe is recognized as one of the fastest computational platforms in proteomics, making it a practical solution for the rapid quality control of high-throughput sample analyses. Starting with version 23.0, FragPipe introduced the “Generate Summary Report” feature, offering .pdf reports with essential quality control metrics to address the challenge of intuitively assessing large-scale proteomics data. While traditional spreadsheet formats (e.g., tsv files) are accessible, the complexity of the data often limits user-friendly interpretation. To further enhance accessibility, PSManalyst, a Shiny-based R application, was developed to process FragPipe output files (psm.tsv, protein.tsv, and combined_protein.tsv) and provide interactive, code-free data visualization. Users can filter peptide-spectrum matches (PSMs) by quality scores, visualize protease cleavage fingerprints as heatmaps and SeqLogos, and access a range of quality control metrics and representations such as peptide length distributions, ion densities, mass errors, and wordclouds for overrepresented peptides. The tool facilitates seamless switching between PSM and protein data visualization, offering insights into protein abundance discrepancies, samplewise similarity metrics, protein coverage, and contaminants evaluation. PSManalyst leverages several R libraries (lsa, vegan, ggfortify, ggseqlogo, wordcloud2, tidyverse, ggpointdensity, and plotly) and runs on Windows, MacOS, and Linux, requiring only a local R setup and an IDE. The app is available at (https://github.com/41ison/PSManalyst.
pubs.acs.org
alisonchaves.bsky.social
It would be amazing if the report.parquet could report the adjacent sequence. Let's say 4 AA before and after the peptide Stripped.Sequence. I do it with left_join from UniProt
id_mapping. This would make the graphical output more complete (as in PSManalyst for FragPipe, github.com/41ison/PSMan...)
alisonchaves.bsky.social
I respect the energy and work you put in the Proteomes. It is not on the good people like you and others I know working for MDPI journals. The editorial group reputation is quite often exposed to this kind of problem.
alisonchaves.bsky.social
Holy crap, what is happening with the undergraduate microbiology courses?
alisonchaves.bsky.social
I can't find the actual publication, but it is quite impressive the level of "don't give a f*" that MDPI gives to peer-review process. Look at the Fig 4 in this one: www.mdpi.com/2079-6382/14... (a mitochondria inside a bacteria!)
alisonchaves.bsky.social
Would be interesting to update that graphic of price over the feature (resolution and scan speed) of interest and the percentage of the sales per manufacturer. I remember Bruker gained traction with timstof family.
alisonchaves.bsky.social
Chomsky is a terrorist now! Stupid people do stupid things as elect Trump.
alisonchaves.bsky.social
Really need to thank Zhou and cols. Great inspiration on how to represent PSM features in protein sequences freely available to the community.
PepMapViz: a versatile toolkit for peptide mapping, visualization, and comparative exploration url: academic.oup.com/bioinformati...
PepMapViz: a versatile toolkit for peptide mapping, visualization, and comparative exploration
AbstractSummary. PepMapViz is a versatile R package that provides flexible peptide mapping and visualization capabilities. PepMapViz can import peptide dat
academic.oup.com
alisonchaves.bsky.social
Only NVIDIA (remember you will need CUDA). The RTX A4500 20 GB GDDR6 4DP seems to be a good option.
alisonchaves.bsky.social
Side note: NVIDIA is not a personal preference. All these algorithms use pytorch, which needs CUDA. Nowadays, there is this ROCm from AMD, but no guarantee it will work. All other specs are flexible. Any 24+ threads 64+ Gb RAM will work. Don't save in GPU if de novo will be routine.
alisonchaves.bsky.social
If you are going to run de novo search, you will need a dedicated GPU. Casanova, Instanovo, Cascadia, and this DIA-BERT will take forever running in CPU. If possible, use an RTX 4000 quadro. Downgrade for cheaper RTX series if it is too expensive. NVIDIA always.
alisonchaves.bsky.social
yeah, gauge of the needle, choice of the anticoagulant, material of the tube (glass or plastic) and several other factors. I guess it is more a matter of be aware of these things, it is hard to standardize this in clinical samples.
alisonchaves.bsky.social
EvoSep is great but quite expensive compared to common nLC. The input is critical for several samples. This alone is enough motivation to choose the nLC system. If people find a way to make microLC useful for real samples, let me know.
alisonchaves.bsky.social
Not convinced at all. Only real difference seems to be at 50 µL/min compared to nL and others flows. Furthermore, you pay the price of huge input of sample. It looks more like walking to the past.
alisonchaves.bsky.social
This reminds me of the case of ubiquitination Degradation Pathway inside of chloroplasts. And I don't know why.
alisonchaves.bsky.social
Feijoada? do we have a Brazilian here? 🤔
alisonchaves.bsky.social
#USP you can do better.
Um edital de concurso público com exigências nada razoáveis. Bastante suspeito!
alisonchaves.bsky.social
Quick check: Is it common for your institutions for the competition announcement to require candidates to collaborate with very specific people (including their names)? Apparently, USP in São Paulo-Brazil has a very specific interest for faculty position as assistant professor.
alisonchaves.bsky.social
The data is pretty new. Check again in a few weeks. It is on the repository side, not the authors or reviewers 🤷🏻
alisonchaves.bsky.social
Yes, it is hard work, but old instruments still can be perfectly used today. We extracted >7k proteins with good quantitative features from an Orbitrap Elite.