Pavol Harar
LightNews LightNews
banner
paloha.bsky.social
Pavol Harar
@paloha.bsky.social
240 followers 560 following 12 posts
Computational Scientist for BioAI & Cryo-Electron Tomography @ ISTA | Co-Founder of https://ACAI.AI.
Posts Replies Media Videos
paloha.bsky.social
Look at this beauty. Light-based hardware for running neural networks efficiently. My totally out-of-expertise idea from years back and scientists at Bioengineering Department, UCLA are making this a reality! www.nature.com/articles/s41...
Optical generative models - Nature
Optical generative models are demonstrated for the rapid and power-efficient creation of never-seen-before images of handwritten digits, fashion products, butterflies, human faces and Van Gogh-style a...
www.nature.com
September 4, 2025 at 12:50 PM
Reposted by Pavol Harar
dzyla.bsky.social
1/5: Over the last few months, I have been working on an update for the #FollowRelionGracefully dashboard, which offers improved real-time job previews for #cryoEM #Relion jobs.
April 27, 2025 at 8:49 PM
Reposted by Pavol Harar
carlbergstrom.com
Let's talk about this Nature piece in more detail.

I've rarely read something so anti-scientific anywhere short of the National Review.

www.nature.com/articles/d41...
Three AI-powered steps to faster, smarter peer review
Tired of spending countless hours on peer reviews? An AI-assisted workflow could help.
www.nature.com
March 6, 2025 at 5:34 AM
Reposted by Pavol Harar
lifeonthewedge.bsky.social
To all #TeamTomo #CryoET and #Chlamydomonas aficionados: we have updated EMPIAR-11830 with a bug fix concerning some cryo-CARE denoised tomos as well as additional files and metadata! 🎉👩🏽‍💻

www.ebi.ac.uk/empiar/EMPIA...

A little thread about what's new... 🧵 1/n
Rare cellular events observed in the cryo-ET dataset. Slices through tomograms (left) with corresponding 3D segmentation (middle and right). Right panels show enlarged views of the boxed regions in the middle panels. A) Mitochondrion at the plasma membrane, tethered to microtubules by thin filaments of unknown identity. Segmented classes: nuclear envelope (light blue), nuclear pore complex (dark blue), 80S ribosomes (orange), mitochondrial membranes (dark red), ATP synthase (royal blue), plasma membranes (grey), microtubules (pink), filaments (yellow), endoplasmic reticulum (purple). Zoomed-in panel shows thin filaments running between microtubules and the mitochondrion at the plasma membrane. Note that this tomogram contains two closely appressed cells, and thus, two plasma membranes with cell walls between. B) Mitochondrial fission event with ER membrane interactions. Segmented classes: mitochondrial membranes (dark red), ATP synthase (royal blue), 80S ribosomes (orange), thylakoid membranes (green), PSII (bright yellow), cell wall (grey), ER (purple), fission site containing filamentous structures perpendicular to the mitochondrial long axis (pale yellow). C) Ciliary transition zone between basal body and axoneme, including assembling IFT train and stellate structure77. Segmented classes: microtubule doublets (light blue), central microtubule pair (light green), stellate structure (pale red), IFT train (yellow). D) Pyrenoid tubule extending from the thylakoids into the phase-separated Rubisco matrix of the pyrenoid. Minitubules originate from thylakoid membranes82. Segmented classes: thylakoid membranes (dark green), PSII (yellow), pyrenoid tubule (lime green), Rubisco (lavender blue), minitubules (orange). Scale bars in A-D: 100 nm. Related to Fig. 2C-H.
February 28, 2025 at 4:35 PM
paloha.bsky.social
Finally, my postdoc work is published in Cell Structure! 🎉 Grateful for the chance to apply deep learning to cryo-ET and learn from @haselbachlab.bsky.social about structural biology🦠. Huge thanks to Lukas Herrmann for coding help and Philipp Grohs for my position and all the GPUs.
impvienna.bsky.social IMP @impvienna.bsky.social · Feb 12
🔬Our @haselbachlab.bsky.social with Pavol Harar have developed ‘FakET’!

The new method creates ‘fake’ electron microscopy images to train AI, reducing manual work in particle identification.

Read the full story: www.imp.ac.at/news/article...

@viennabiocenter.bsky.social

#AI #cryoem #microscopy
February 15, 2025 at 10:08 AM
Reposted by Pavol Harar
cellarchlab.com
WOW!😍 This is exactly the type of development we hoped the Chlamy dataset would help empower! Lots of organelles and cellular features labeled automatically, context-aware particle picking, & area-selective template matching #TeamTomo 🔬

Big props to @mgflast.bsky.social @thomsharp.bsky.social 🧪🧶🧬
aixbiobot.bsky.social AI x Bio Discovery @aixbiobot.bsky.social · Jan 18
Scaling data analyses in cellular cryoET using comprehensive segmentation [new]
Comprehensive segmentation of cryoET data enables ontology-based voxel ID, used for context-aware particle picking and faster template matching.
Scaling data analyses in cellular cryoET using comprehensive segmentation Figure 1 Figure 2 Figure 3
January 19, 2025 at 9:04 PM
paloha.bsky.social
Thanks Alicia! I am also happy to have helped finalize this work. I hope computational biologists will find the repo useful. We need to share more annotations in accessible format, the algorithms are hungry 😄 #teamtomo, consider sharing your data via pull requests—let’s grow this together! 🤝
aliciakmichael.bsky.social Alicia Michael @aliciakmichael.bsky.social · Jan 6
I am happy to be a part of this effort and looking forward to the discoveries this dataset will enable! If you work with this dataset in the future, please add any annotations to the GitHub repository github.com/Chromatin-St... Thanks @paloha.bsky.social for spearheading the efforts with this!
January 7, 2025 at 1:18 PM