This work wouldn't have been possible without the fantastic collaboration of Anant Nawalgaria from #GoogleAI , the mentorship of Matthias Mann, and the incredible support of the whole team at @mannlab.bsky.social‬. A huge thank you to everyone involved! #Science #Collaboration

Under the hood, our agentic framework is built on Google's Agent Development Kit (ADK). We leveraged Gemini 2.5 pro for the complex video analysis and Gemini 2.5 flash for the main agent functionalities, all using Google Cloud infrastructure. #GoogleAI #Gemini #MultimodalAI 6/🧵

Our goal is to enhance reproducibility and to make cutting-edge science more accessible. We believe this happens when AI is designed to augment - rather than replace - the rigor and intuition of a scientist. To this end, our code is open-source on GitHub: github.com/MannLabs/pro... 5/🧵

📈 Instrument guidance: Is the instrument ready for measurement? Our agent answers by benchmarking current performance against historical QC data and their related decisions. It also preserves the final troubleshooting reasoning to build institutional memory. #Proteomics 4/🧵

🔍 AI-assisted error detection: By comparing a video of an experiment to the baseline protocol, our agent detects common mistakes. It found 3/4 of all errors in our benchmark dataset, though domain-specific knowledge & spatial recognition need improvement. #GenerativeAI #QualityControl 3/🧵

✍️ Automated protocol generation: Just record yourself doing an experiment while explaining the steps. Our agent analyzes the video and audio to generate a detailed, formatted protocol in minutes. #MultimodalAI #FutureOfScience 2/🧵

In labs, hands-on expertise is often lost because it's not written down. We leverage multimodal AI agents to capture & share expertise by analyzing video and speech to generate protocols, detect errors, and guide researchers. #AI #TeamMassSpec
📄 Preprint: doi.org/10.1101/2025... 1/🧵

Thank you! To run synchro-PASEF, you need the hardware of a timsTOF HT or timsTOF Ultra (older instruments can be upgraded) & timsControl 6

Looking for rapid, deep & reproducible proteomics? Follow our @Nature protocol to get 7k proteins & 29k phosphosites in just 21 min! Complete #PASEF workflow using optimal dia-PASEF, synchro-PASEF & py_diAID. From method development to analysis. #TeamMassSpec @patiskowronek.bsky.social explains 👇

Here is the correct link: doi.org/10.1038/s415...

Team Work: Most important, this work published in Nature protocols is the result of an amazing team. Special thanks to my brilliant colleagues: @georgwa.bsky.social @mariawahle.bsky.social @swillems.bsky.social @mannlab.bsky.social @mpibiochem.bsky.social and to #Bruker for their great support. 11/🧵

Anticipated Results: Expect deep coverage with 7,000 protein groups and 29,000 Class I phosphosites in 21 min! Synchro-PASEF combines high sensitivity, selectivity and speed - no compromises. Shows superior quantitative performance, especially at low inputs and fast gradients. #proteomics 10/🧵

🔧 Troubleshooting: We have you covered for (every) challenge you encounter. We explain how to the clean the emitter, improve timsTOF performance, or manage sample load. #troubleshooting #proteomics #TeamMassSpec 9/🧵

📋 Step-by-step #protocol of PASEF workflow: From calibration to automated method generation with py_diAID. Dive into #timsTOF acquisition followed by analysis with AlphaDIA, Spectronaut & DIA-NN. Use our dia-PASEF & synchro-PASEF methods. Everything you need for successful #proteomics analysis! 8/🧵

Methods selection simplified: One robust method covers all standard proteomics samples. For modified peptides like #phosphoproteomics #peptidomics #HLA specialized methods (orange) are essential. Why? Chemical properties dictate precursor behavior. 7/🧵

Fine-tuning PASEF:
1) Optimal DIA isolation windows maximize precursor cloud coverage.
2) Short cycle times (500ms with synchro-PASEF) ensure high reproducibility.
3) Need enhanced sensitivity? Longer TIMS accumulation captures low abundant ions.
#proteomics #massspec 6/🧵

🛠️ Meet py_diAID on GitHub: Our open-source #python tool that makes dia-PASEF & synchro-PASEF method development easy! Optimize your windows for any sample type - from #proteomics over #phosphoproteomics to #peptidomics #HLA. Validate methods before running or transform your own method into a GIF 5/🧵

⚡️ synchro-PASEF: Our newest scan mode uses a super-fast moving quadrupole for high selectivity with up to precursor resolution - keeping all advantages of DDA & DIA. Result: 2-3x faster cycle times than dia-PASEF (only 500ms) AND superior quantitative reproducibility. #synchroPASEF #DIA 4/🧵

dia-PASEF: Optimal DIA isolation windows in both m/z AND ion mobility space maximize precursor coverage. Our optimized methods capture nearly the entire precursor cloud in shorter cycle times - giving you a robust method with deep proteomes. #diaPASEF #DIA 3/🧵

#PASEF 's magic: By decoupling ion accumulation from mass detection, PASEF captures nearly ALL ions with high sensitivity. Plus, ion mobility separation gives you cleaner spectra by sorting precursors & fragments in an extra dimension. Less overlapping peaks! #timsTOF #Bruker 2/🧵

🚀 Robust and high sensitivity #proteomics: Our Nature protocol demystifies #PASEF workflows and provides ready-to-use dia-PASEF & synchro-PASEF methods. Find out how to achieve 7,000 protein groups or 29,000 phosphosites in 21min. Let's explore! #TeamMassSpec #Bruker doi.org/10.1038/s415... 1/🧵