Dictates to the media.
Mentor, scientist & engineer. Having fun in @slavovlab.bsky.social and Parallel Squared Technology Institute @parallelsq.bsky.social with biology & single-cell proteomics. https://nikolai.slavovlab.net
2,297 followers 132 following 631 posts
view profile on Bluesky
Dictates to the media.
Compression causes cancer cells to start spreading www.nature.com/articles/d41...
Mutations play important roles in cancer. They are one important factor among many others leading to uncontrolled cell proliferation and metastasis. Cancer is a disease involving genetic, signaling, metabolic, endocrine, and other abnormalities.
These findings add to the increasing evidence that cancer is not driven solely by DNA mutations.
The biomass distribution on Earth www.pnas.org/doi/abs/10.1...
Biomass distributions across different environments and trophic modes.
Surprising estimates about the biomass distribution on Earth: ◼️Terrestrial biomass is ~ 100-fold higher than marine biomass ◼️ Plans are mainly terrestrial, whereas animals are mainly marine ◼️ Live stock biomass exceeds human biomass
Many biological processes can be thought of as the result of an underlying dynamics in which the system repeatedly undergoes distinct and abortive trajectories with the dynamical process only ending when some specific process, purpose, structure or function is achieved. arxiv.org/abs/2506.04104
Life uses exploratory dynamics. It's a simple & powerful approach: ◼️ Variation followed by selection 1/2
I excavated and reviewed the supplementary figures. They were very disappointing.
I am reading a research article, and the key results are shown in supplemental figures, even though the authors used fewer main figures than the journal allows. Strange.
The word 𝑤𝑟𝑖𝑡𝑖𝑛𝑔 may mislead as I believe the focus should be on 𝑡ℎ𝑖𝑛𝑘𝑖𝑛𝑔, 𝑓𝑟𝑎𝑚𝑖𝑛𝑔, 𝑎𝑛𝑑 𝑜𝑟𝑔𝑎𝑛𝑖𝑧𝑎𝑡𝑖𝑜𝑛. Communicating a significant new scientific result is a highly intellectual process with many subtleties. 2/2 blog.slavovlab.net/2024/02/12/c...
A timeless model for writing a research summary. 1/2
Proteins self-organize to build large structures, 100,000-fold larger then the proteins. It's elegantly dynamic ! Proteins constantly self-assemble and disassemble to build skeletal structures in our cells, shaping and orchestrating life.
... we should also acknowledge the benefits and the ongoing increase in transparency. 2/2 bsky.app/profile/slav...
This summer we received numerous thoughtful and constructive reviews. I am glad more journals now publish peer reviews and look forward to their publication. Peer review remains highly imperfect, but ... 1/2 blog.slavovlab.net/2023/04/15/p...
We share our technologies to maximize the benefit for the community. PSMtags enable 9-plexDIA with better sequencing and multiplexing. PSMtags support sensitive proteomic analysis of over 1,000 samples / day. parallelsq.org/psmtags
The end of the genetic paradigm of cancer journals.plos.org/plosbiology/...
Mutations play important roles in cancer. They are one important factor among many others leading to uncontrolled cell proliferation and metastasis. Cancer is a disease involving genetic, signaling, metabolic, endocrine, and other abnormalities.
This provocative article does not mince words. The authors argue that genome sequencing and scRNA-seq yield findings challenging the notion that cancer is a ‘genetic disease’: "... the cancer research community has abandoned deep thinking for deep sequencing" 1/n
Sensitive mass spec requires efficient ion utilization. ◼️Slice-PASEF increases this efficiency and thus the sensitivity of MS. We demonstrate its benefits by analyzing the proteomes of single lymphocytes, which reflect benefits observed in other @slavovlab.bsky.social projects using Slice-PASEF.
A gene may template many dozens of products. Its products can have diverse, even opposing functions. Ascribing the functions of the products to the gene is a simplistic approximation that has outlived its usefulness.
The past decade has been full of exciting ideas, datasets, and discoveries. I am proud of our alumni who now drive innovation in industry, lead their own research groups, and spread a culture of bold and rigorous research. Excited for what the next decade brings.
An essential component of a virtual cell model.
This description strongly resonates with me when I think of our first mass spectra from peptides derived from single cells. The excerpt is from: "Orbitrap Against All Odds" theanalyticalscientist.com/issues/2013/...
𝐌𝐚𝐣𝐨𝐫 𝐫𝐞𝐬𝐮𝐥𝐭𝐬 𝐚𝐫𝐞 𝐫𝐚𝐫𝐞𝐥𝐲 𝐛𝐨𝐫𝐧 𝐩𝐞𝐫𝐟𝐞𝐜𝐭. Here's a breakthrough story in the words of the lead innovator: 𝑇ℎ𝑒𝑦 𝑤𝑒𝑟𝑒 𝑝𝑟𝑒𝑡𝑡𝑦 𝑠𝑖𝑚𝑝𝑙𝑒 𝑑𝑎𝑡𝑎 𝑓𝑟𝑜𝑚 𝑢𝑛𝑟𝑒𝑙𝑖𝑎𝑏𝑙𝑒 𝑎𝑛𝑑 𝑢𝑛𝑠𝑡𝑎𝑏𝑙𝑒 𝑒𝑞𝑢𝑖𝑝𝑚𝑒𝑛𝑡, 𝑏𝑢𝑡 𝑡ℎ𝑒𝑦 𝑠ℎ𝑜𝑤𝑒𝑑 𝑜𝑓𝑓 𝑡ℎ𝑒 𝑢𝑛𝑚𝑖𝑠𝑡𝑎𝑘𝑎𝑏𝑙𝑦 ℎ𝑖𝑔ℎ 𝑟𝑒𝑠𝑜𝑙𝑣𝑖𝑛𝑔 𝑝𝑜𝑤𝑒𝑟 𝑜𝑓 𝑡ℎ𝑒 “𝑂𝑟𝑏𝑖𝑡𝑟𝑎𝑝” -- 𝐀𝐥𝐞𝐱𝐚𝐧𝐝𝐞𝐫 𝐌𝐚𝐤𝐚𝐫𝐨𝐯
I started a Substack blog discussing a favorite topic: 𝐖𝐡𝐢𝐜𝐡 𝐩𝐫𝐨𝐣𝐞𝐜𝐭𝐬 𝐝𝐞𝐬𝐞𝐫𝐯𝐞 𝐨𝐮𝐫 𝐥𝐢𝐦𝐢𝐭𝐞𝐝 𝐭𝐢𝐦𝐞 ? open.substack.com/pub/nikolais...
The total mass, number, and distribution of immune cells in the human body www.pnas.org/doi/10.1073/...
Lymphocytes: ~ 40% of the number, and 15% of the mass Neutrophils: ~ 40% of the number, and 15% of the mass Macrophages: ~ 10% of the number, and 50% of the mass
The distribution of ~ 2 trillion immune cells in the human body.
Our findings (now published in @narjournal.bsky.social) identify the first metazoan RNase MRP-specific protein subunits. We also define the RNA-targeting repertoire of this essential enzyme in mammalian cells. academic.oup.com/nar/article/...
The key molecules in immune responses are glycoproteins. Why isn't glycosylation studied more intensively? It should be. It's a challenging research area, but a good one for ambitious & audacious scientists.
Congratulations 🙌 It is richly deserved !
Mitochondria have many functions beyond their well-established bioenergetic and biosynthetic functions. They integrate physiologic inputs and communicate signals that control cell function or fate as well as adaptation to stress.
Tools have long propelled discoveries: "𝑁𝑜𝑡ℎ𝑖𝑛𝑔 𝑡𝑒𝑛𝑑𝑠 𝑠𝑜 𝑚𝑢𝑐ℎ 𝑡𝑜 𝑡ℎ𝑒 𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑚𝑒𝑛𝑡 𝑜𝑓 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 𝑎𝑠 𝑡ℎ𝑒 𝑎𝑝𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝑎 𝑛𝑒𝑤 𝑖𝑛𝑠𝑡𝑟𝑢𝑚𝑒𝑛𝑡" Sir Humphry Davy Elements of Chemical Philosophy (1812)
The registration for SCP2026 is open ! Join us next July for the the 9th Single Cell Proteomics Conference. single-cell.net/proteomics/s...
Scientists should actively promote good research. 𝐖𝐞 𝐧𝐞𝐞𝐝 𝐭𝐨 𝐝𝐨 𝐢𝐭 𝐭𝐡𝐨𝐮𝐠𝐡𝐭𝐟𝐮𝐥𝐥𝐲, 𝐜𝐫𝐢𝐭𝐢𝐜𝐚𝐥𝐥𝐲 𝐚𝐧𝐝 𝐜𝐨𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐯𝐞𝐥𝐲. blog.slavovlab.net/2020/07/15/p...
Single-cell differences in matrix gene expression do not predict matrix deposition www.nature.com/articles/nco...
Marker gene expression is a poor predictor of cartilage-like matrix production in individual mesenchymal stem cells.
If you measure and interpret RNA levels in single cells, read this: "𝑂𝑢𝑟 𝑟𝑒𝑠𝑢𝑙𝑡𝑠 𝑠𝑢𝑔𝑔𝑒𝑠𝑡 𝑡ℎ𝑎𝑡 𝑖𝑛𝑠𝑡𝑎𝑛𝑡𝑎𝑛𝑒𝑜𝑢𝑠 𝑚𝑅𝑁𝐴 𝑎𝑏𝑢𝑛𝑑𝑎𝑛𝑐𝑒 𝑜𝑓 𝑐𝑎𝑛𝑜𝑛𝑖𝑐𝑎𝑙 𝑚𝑎𝑟𝑘𝑒𝑟𝑠 𝑖𝑠 𝑡𝑒𝑛𝑢𝑜𝑢𝑠𝑙𝑦 𝑙𝑖𝑛𝑘𝑒𝑑 𝑡𝑜 𝑡ℎ𝑒 𝑐ℎ𝑜𝑛𝑑𝑟𝑜𝑔𝑒𝑛𝑖𝑐 𝑝ℎ𝑒𝑛𝑜𝑡𝑦𝑝𝑒 𝑎𝑡 𝑡ℎ𝑒 𝑠𝑖𝑛𝑔𝑙𝑒-𝑐𝑒𝑙𝑙 𝑙𝑒𝑣𝑒𝑙." 1/2
When we started this project - about 10 years ago - we had limited access to old equipment. Better equipment would have been helpful, but using older equipment changed the perception about the feasibility & accessibility of single-cell proteomics. genomebiology.biomedcentral.com/articles/10....
We mostly ask different questions now. Can we do better ? Can we formulate more creative hypotheses and launch more creative investigations ? Many fundamental questions remain open.
Such simple and effective approaches are not common in modern biomedical research. Why ?
This article is a confluence of classical genetics & biochemistry that spearheaded understanding the secretory pathway. It used an ingenious sedimentation method to isolate mutations in genes coding for protein from the secretory pathway. Brilliantly simple. Powerful❗️ 🧵
The epitranscriptome formed by the growing number of modifications occurring within mRNA transcripts.
Protein degradation is often considered as a final step in regulating gene expression. It also regulates the first step: Transcription Many TFs, including p53, HIF-1α and c-Myc, are regulated by 𝐝𝐞𝐠𝐫𝐚𝐝𝐚𝐭𝐢𝐨𝐧-𝐠𝐚𝐭𝐞𝐝 𝐜𝐨𝐧𝐭𝐫𝐨𝐥: ⬛️ It has a key advantage. 👇
Nucleoside diphosphate kinase A (NME1) catalyses its own oligophosphorylation www.nature.com/articles/s41...
Biology is full of surprises and open questions. blog.slavovlab.net/2024/09/02/f...
Enter protein oligophosphorylation. A protein catalyses its own oligophosphorylation, up to the formation of a hexaphosphate chain. That is cool !
Neighboring gene effects account for much of the observed phenotypes of gene deletions. This is a crucial factor to consider in knockout experiments. Phenotypic similarity is exponentially related to chromosomal proximity in both yeast & human genomes. Human Data 🔽
Richard Nixon's administration attempted to reduce, eliminate or impound funding for research and training grants. cancerletter.com/the-cancer-l...
Protein synthesis is among the most energy demanding processes in a cell. 𝐖𝐡𝐲 𝐚𝐫𝐞 𝐭𝐡𝐞𝐧 𝐬𝐨𝐦𝐞 𝐩𝐫𝐨𝐭𝐞𝐢𝐧𝐬 𝐜𝐨𝐧𝐭𝐢𝐧𝐮𝐨𝐮𝐬𝐥𝐲 𝐬𝐲𝐧𝐭𝐡𝐞𝐬𝐢𝐳𝐞𝐝 𝐚𝐧𝐝 𝐢𝐦𝐦𝐞𝐝𝐢𝐚𝐭𝐞𝐥𝐲 𝐝𝐞𝐠𝐫𝐚𝐝𝐞𝐝 ? blog.slavovlab.net/2025/08/16/p...
Is single-molecule protein sequencing here yet? 2/2 www.nature.com/articles/s41...
The answer is simple: 🔳 No method can sequence a protein molecule. Some methods can fingerprint protein molecules. The complexity and scale of the proteome is a challenge for all methods, especially for single-molecule ones: nature.com/articles/s41... 1/2
... and thus having different distributions across the phases of the cell division cycle (CDC) and yeast metabolic cycle (YMC). The growth trends in abundance (GR slopes) of CDC-periodic mRNAs are explained by the YMC but not the CDC cycling dynamics. www.pnas.org/doi/full/10....
The fraction of cells in each phase should be proportional to the relative duration of the phase. We applied this model to cells growing at different growth rates, and .... www.molbiolcell.org/doi/10.1091/... www.pnas.org/doi/full/10....
I collected these data ~ 15 years ago, but we still do not understand puzzling observations. ⬛️ mRNAs whose abundance is periodic with the phases of the cell division cycle (CDC) do not exhibit the growth rate trends expected from the CDC. 🧵 www.molbiolcell.org/doi/10.1091/...
AI content is tainting research publications www.nature.com/articles/d41...
Stating our results, framed the way we want them framed, is part of the fun and privilege of communicating research. I would not share this fun (and responsibility) with a chatbot.
AI companies brag about “huge demand” for AI subs. Demand is huge only at heavily subsidized prices. VCs pay the bills. Nobody knows what demand will be at sustainable prices because the data doesn’t exist.
Other factors, noted by the authors, are: Genetic associations reflect a culture-specific interplay between underlying heritable factors and environmental influences. Polygenic indices do not necessarily yield accurate predictions at the level of the individual. www.cell.com/current-biol...
Either most modern humans are musically more gifted than Beethoven was or the polygenic predictions are unreliable. The second possibility is much more likely. Indeed, the polygenic indices in this case (and more generally) account for only a small fraction of the phenotypic variance.
Researchers used polygenic indices to predict Beethoven’s musical propensity. The polygenic predictions ranked Beethoven in the bottom percentile of two moderns cohorts. Why? 🧵
Whatever drives this trend, it’s not a deep understanding of biology. Biology has many unexplored frontiers: blog.slavovlab.net/2024/09/02/f...
Cherish your competitors: They understand your mission. They prove your goals are worth chasing.
Cell line heterogeneity is not unique to HeLa, though recently derived cell lines likely have less variability across labs. Many cell lines have less genome abnormalities and were acquired with informed consent.
A good example of quantifying multiple distinct contributors to transcriptional regulation, including TF activation domains, binding affinity, binding site configuration and recruitment of chromatin regulators. www.nature.com/articles/s41...
Julia Schaepe presented these results at the 8th single-cell proteomics conference: They include simple thermodynamic & kinetic models with good predictive power. youtu.be/B06OeSeaMgg?...
An elegant model: - TF binding events on nucleosome-free DNA are independent. - Yet, the activation domains of TFs recruit cofactors that destabilize nucleosomes, resulting in cooperative TF binding. - Average TF occupancy linearly determines promoter activity
I am often asked if we perform 𝑢𝑙𝑡𝑟𝑎𝑠𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑒 𝑝𝑟𝑜𝑡𝑒𝑜𝑚𝑖𝑐𝑠. I don't know what to reply. I'm not aware of a definition of "𝑢𝑙𝑡𝑟𝑎𝑠𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑒 𝑝𝑟𝑜𝑡𝑒𝑜𝑚𝑖𝑐𝑠". As discussed in this editorial, "𝑢𝑙𝑡𝑟𝑎𝑠𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑒" is confusing. I prefer precise terminology.
Phenomenological associations can suggest biomarkers. The significance and interpretation of such biomarkers are limited without some understanding of the underlying molecular mechanisms.
The heterogeneity of HeLa cell-lines undermines reproducibility. HeLa cells were obtained without informed consent. ==> Researchers should avoid using HeLa cells unless required by a special justification.
Science needs more public discussions focussed on constructive criticism. Such discussions are hard. Much harder than superficial praise that is popular on social media. Yet, scrutiny & rigor are more important for science than superficial praise.
Consider the other side
Quantitative protein assays on Illumina chips. A creative use of the impressive imaging capabilities developed for DNA sequencing by synthesis. ◼️ It allowed quantifying binding and enzyme catalysis 🧵
RNA-seq is usually DNA-seq. This has consequences: ◼️ The indirect readouts of RNA by sequencing complementary DNA usually lack information about RNA modifications.
This article posits that AI models can derive fundamental biological principles. What is an example of a fundamental biological principle derived by an AI model ? ?
Pan-cancer analysis of whole genomes www.nature.com/articles/s41...
No driver mutations could be identified in 181 tumors. The authors do not know why. Perhaps, it's an algorithmic failure. Perhaps, it reflects non genetic contributions to carcinogenesis. What do you think ?
Billions of💲 are invested in developing treatments promoting longevity & happiness: The most effective approach is free: - Physical activity. Move and enjoy life !
Formins are engines powering cellular acrobatics. They are Rho family GTPases that orchestrate cytoskeleton rearrangements. They can nucleate the polymerization of actin filaments, accelerate the elongation of actin filaments, and stabilize microtubules.
Building meaningful & generalizable representations of biological systems is very challenging, especially at the current state of our understanding. Here is what I think is required: www.youtube.com/watch?v=P0-_...
The datasets used did not include key regulators. Thus, the models either ignore them completely or have to model them as latent variables. ⬛️ Both approaches lead to models based on indirect correlations with a low probability of generalizing beyond training distributions.
One factor is the 𝐬𝐢𝐳𝐞 of the training data. - It gets much attention. Another factor is the 𝐭𝐲𝐩𝐞 of the training data. - It gets less attention. 1/n
This interview offers an accessible and intuitive overview of his estimates: . youtu.be/-zF1mkBpyf4?...
Daron Acemoglu expects AI to contribute "no more than a 0.71% increase in total factor productivity over 10 years." economics.mit.edu/sites/defaul...
A simple benchmark. Most people can understand it. No PhD-level intelligence needed.
Widespread Post-transcriptional Attenuation of Genomic Copy-Number Variation in Cancer www.cell.com/cell-systems...
Widespread Post-transcriptional Attenuation of Genomic Copy-Number Variation in Cancer
A significant contribution to this buffering stems from proteins interactions resulting in protein complexes. Subunits of protein complexes that do not become incorporated in a complex are often degraded. Protein degradation is a major regulator of protein abundance doi.org/10.1101/2025...
DNA copy number variation is strongly buffered at the protein level. The buffering is much stronger than at the RNA level.
In addition to the scale, the type of the data matters too. I think datasets tend to miss key regulators. Inference is likely limited by confounders and the fact that the regulators are latent variables not even included in the models. blog.slavovlab.net/2019/04/07/d...
Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines www.nature.com/articles/s41...
The authors report that all seven tested deep learning models failed to outperform simple baselines in predicting transcriptome changes after single or double perturbations. I will duck for cover.
There is no simple way to evaluate universities. Based on reputations, many universities seem to be among the organizations sustaining success over longer time scales.
Few successful organizations sustain excellence for decades. Even fewer for centuries. This is reflected in the perpetual flux in the top 10 companies by market cap - an accessible, albeit not great, measure of success. ⬇️ Top companies in the Fortune 500 list, circa 1955
Mechanotransduction in the nucleus Multiple TFs transmit mechanical forces into transcriptional regulation. Nuclear shape, size, and chromatin abnormalities are common features of many diseases.
How to Scale Proteomics press.asimov.com/articles/pro...
Scaling DNA sequencing is a notable success story. Scaling protein analysis is harder since the proteome is a much taller mountain: www.nature.com/articles/s41... @asimovpress.bsky.social writes about @parallelsq.bsky.social's approach to scale up proteomics.
