Information is more accessible than ever. AI models excel at information retrieval. And yet, the quality of popular & influential information sources is declining. How do we reverse this dangerous trend ? ℹ️

Developing treatments for Alzheimer’s and Parkinson’s diseases is limited by the incomplete understanding of the molecular processes underlying them. Single-cell proteomics is opening a new window into these diseases: - It allows for a different type of analysis, different in kind. 1/

We're looking for mass spectrometry enthusiasts who thrive on quantitative rigor, innovation and enjoy challenges: - Colleagues eager to propel biological discoveries by inventing the technologies of the future.

A pioneer of AI-in-biology rebuts the "AI will cure cancer" narrative. Not smarter models. Measurements. 90%+ of clinical trials fail — usually the molecule was fine, the mechanism was wrong. Scalable mass spec proteomics = a real path to the functional data we're missing. 🧵

I enjoyed this conference ! Many presentations combined relatively simple, intuitive biophysical models with clear predictions that were experimentally tested. I especially appreciated when speakers were explicit about their study assumptions and limitations, facilitating critical discussions.

I recently attended two Gordon Research Conferences on very different topics. The most memorable talks had one thing in common: They focused deeply on a single project, taught something new with rigor, made their assumptions and caveats clear, and invited critical thinking. 1/

One fold. Many different jobs. Antibodies, T-cell receptors, MHC molecules, cell adhesion proteins, even a giant muscle protein like titin — all of them use the same ancient structural building block: a small β-sandwich domain called the immunoglobulin fold. 1/

America's scientific leadership was built on Vannevar Bush's vision that investing in fundamental research creates tomorrow's breakthroughs. Abandoning this vision risks weakening the engine that has driven U.S. prosperity, innovation, and health for generations. 1/

Dimensionality reduction can see structures that do not exist and miss structures that exist. 𝐓𝐡𝐞 𝐬𝐢𝐦𝐩𝐥𝐞𝐬𝐭 𝐞𝐱𝐩𝐥𝐚𝐧𝐚𝐭𝐢𝐨𝐧 𝐢𝐬𝐧’𝐭 𝐚𝐥𝐰𝐚𝐲𝐬 𝐭𝐡𝐞 𝐛𝐞𝐬𝐭 𝐨𝐧𝐞. Sometimes, no simple way of describing the complexity of the data exists, at least not one that humans can readily interpret. 1/

Biology drowns in data, yet starves for measurement. Most biological molecules have never been measured; those that have were often captured only under narrow, limited conditions. What we collect are snapshots, not movies. Static frames of systems that are fundamentally dynamic. 1/

This Nature Chemical Biology article highlights how alternate RNA decoding is expanding the proteome diversity. The highlighted results are reshaping my view of the human proteome and understanding of biology. www.nature.com/articles/s41...

Communicating research results is among the most important aspects of science; it can be tremendously rewarding and gratifying. Yet, it often falls short. This is an insightful & humorous take on giving a talk by @itaiyanai.bsky.social. 1/

Thinking on the timescale of days & weeks, progress can feel frustratingly slow. Shifting to the timescale of years, the progress is spectacular: Often faster than expected. The arc of technology development has been awesome. The best is yet to come 🚀 annualreviews.org/content/jour...

Some proteins covary with the cell division cycle similarly across cell types. Others covary in a cell type-dependent way. What is your interpretation ? doi.org/10.1186/s130...

Appreciating the unknown helps us discover it. Great oceans of truth lay all undiscovered by modern science. One of them is them is the proteome: nikolai.slavovlab.net/Proteome-sec...

Paying peer reviewers led to faster first editorial decisions — an average of 5.5 working days, down from nearly 38 for unpaid reviews. The review quality, as judged by handling editors, went up. Do you see this model changing peer reviewer across the larger ecosystem ? 1/

Identifying the processes contributing to the measured abundance of substituted peptides (RNA codons, modifications and proteoform degradation rates) further increases the confidence in the estimated abundances. 1/7

Detecting the same amino acid substitutions in peptides produced from different proteases provides further confidence. 1/6

Deep learning can predict peptide fragmentation spectra and chromatographic properties with high accuracy and precision. These predictions become are useful validation methods, which allow to rescore identification confidence (PEP; posterior error probability) 1/5

What is more difficult is the validation. Building enough confidence required cross validation by many methods and took years. An inflection point in the cross-validation was triangulating across all peptides from proteoforms and finding consistent support for AAS ratios. 4/

Our recent Nature paper expands the usual set of hypotheses to peptides differing by a single amino acid from the predictions of the genetic code. This can be done systematically and is the easier part. 3/

A deeper dive into the article we published last week in @nature.com. I will start with the framing and methodology as they are key to understanding the surprising results: Most proteomics experiments begin with an assumption. www.nature.com/articles/s41... 1/

As usually, the Single-Cell Proteomics Conference aims to maximize accessibility. Thus, #SCP2026 will enable virtual attendance via Zoom. It's free but requires registration at single-cell.net Only presentations by speakers who consent to Zoom broadcast will be included in the virtual program.

Alternate RNA decoding is pervasive across functional groups of proteins, healthy and diseased tissues. It affects proteins playing key roles in neurodegeneration, and some alternately decoded proteins show strong enrichment in tumors compared to their surrounding tissues.

These proteins are not rare translation byproducts. They accumulate to thousands of copies per cell. Some are more abundant than the proteins predicted by the genetic code from the same transcripts.

Since the 1960s, the genetic code has been used to predict protein sequences from DNA and mRNA sequences. Our @nature.com article demonstrates that these predictions miss thousands of protein sequences present in human tissues. www.nature.com/articles/s41... 1/

A decade ago, the money wasted on poor antibodies was estimated to be US $800 million / year. Global research antibody spending is ~$4-6B/year now vs $1.6B in 2015. What is your estimate for money wasted on poor antibodies in 2026 ?

For an elite group of capable, ambitious, and highly motivated scientists, a great postdoc remains one of the most powerful career accelerators available. The challenge is ensuring that more postdoctoral positions function as training for scientific leadership.

Multiple analyses evaluate postdoctoral positions in financial terms. But discussions focused narrowly on average earnings are incomplete. Salary is easy to quantify, especially in the shorter term: Not everything that can be counted counts, and not everything that counts can be counted. 1/

The impact factor (IF) is a limited metric. It's flawed. Yet, it's influential. Less influential than many think, but more influential than it should be. It's up to us -- the scientific community -- to avoid undue reliance on limited metrics and support broad and rigorous research evaluation.