AI ‘n Data @Deloitte; Ex @OchreBio @AstraZeneca @GenomicsEngland @imperialcollege @NCSR_Demokritos; Views my own; RT != endorsement
Cambridge, England
Joined May 2016
- Tweets1.8K
- Following534
- Followers648
- Likes1.5K
D.Polychronopoulos retweeted
Our latest paper has just been published in Cell!
doi.org/10.1016/j.cell.2025.…
We developed a new method called MCC ultra, which allows 3D chromatin structure to be visualised with a 1 base pair pixel size.
D.Polychronopoulos retweeted
Just in time for my 29th birthday today, I am so excited to present: gget v0.29.0 🎉
We added four new modules that are all about
- gene-disease effects
- cancer
- drugs
- therapeutic targets
🧵
D.Polychronopoulos retweeted
It's great to see my first paper of the PhD published in @NatureBiotech! go.nature.com/47ARizC. This describes our scRNA-seq method, scITD, for analyzing how the transcriptional states of multiple cell types are linked across patients. A few highlights from the paper:
D.Polychronopoulos retweeted
The in-print version of the PathChat @Nature article is now available online with open access nature.com/articles/s41586-0… learn more about next steps and how PathChat is further being developed at @modella_ai
D.Polychronopoulos retweeted
Interested in applying Prophet for your own functional assay? Check out the manuscript biorxiv.org/content/10.1101/… or grab the model itself: github.com/theislab/prophet
Really excited about this new predictive approach for phenotype & drug discovery, check it out!
D.Polychronopoulos retweeted
Does the brain change during pregnancy? You bet it does. How about doing 26 MRI scans on a single person pre-conception to 2 years post-partum. WOW! Laura Pritschet and colleagues nailed it with their new paper in @NatureNeuro.
Key Points:
- The authors point out that pregnancy is a period of profound hormonal and physiological changes.
- The authors did something incredible by mapping neuroanatomical changes in an individual from preconception through two years postpartum.
- Boom! There were pronounced decreases in gray matter volume and cortical thickness.
- There were increases in white matter microstructural integrity, ventricle volume and cerebrospinal fluid.
- There were few regions untouched by the 'transition to motherhood.'
My take: This study shows us how a single brain in a single person can be a powerful resource when scanned serially over time. This method is hard to pull off, however what it has the potential to teach us about neuroscience is mind-blowing. Though this paper is only a single subject, the data provides a comprehensive map of the human brain across gestation. It will be interesting to add to the dataset, so we can understand changes across multiple persons and multiple conditions. The convergence of evidence is revealing that pregnancy as associated with dramatic neuroplasticity. How about a shout out to this 38-year-old woman who underwent 26 MRI scans along with blood draws from 3 weeks preconception through 2 years postpartum!
nature.com/articles/s41593-0… #pregnancy #brain @ndosenbach @andreashorn_ @EMiddlebrooksMD
D.Polychronopoulos retweeted
Matthias @labs_mann shows first joint proteome and RNA workflow in same cell - very exciting, and lots of promises to build multimodal atlases in nearer future!
D.Polychronopoulos retweeted
Happy that our ehrapy toolbox for the exploratory analysis of electronic health records is out today @NatureMedicine! It enables early QC&imputation at scale, visualization and downstream clustering & patient trajectory learning.
nature.com/articles/s41591-0… github.com/theislab/ehrapy
D.Polychronopoulos retweeted
The field of #ComputationalBiology has changed dramatically in the last few years with #GenerativeAI.
I'm excited to be recording my lectures again this year, and to be teaching with the awesome @EjAlm.
You can find the first lecture here: youtu.be/1zZSPeKGRzw
and more will be coming in the playlist below.
Fall 2024: youtube.com/playlist?list=PL…
Spring 2021: youtube.com/watch?v=0jWOZoTs…
Fall 2020: youtube.com/watch?v=CTPs5HEL…
#CompBio #DeepLearning #MachineLearning #Genomics #Therapeutics #RepresenationLearning #Chemistry #ProteinFolding #ClinicalRecords #SingleCell
Today was my last lecture for @MIT #ComputationalBiology: #Genomes, #Networks, #Evolution, #Health. I recorded each and immediately posted online here: youtube.com/playlist?list=PL…
Please do share, and let me know which topics need more explanations, clarifications, and corrections!
D.Polychronopoulos retweeted
Programming is changing so fast... I'm trying VS Code Cursor + Sonnet 3.5 instead of GitHub Copilot again and I think it's now a net win. Just empirically, over the last few days most of my "programming" is now writing English (prompting and then reviewing and editing the generated diffs), and doing a bit of "half-coding" where you write the first chunk of the code you'd like, maybe comment it a bit so the LLM knows what the plan is, and then tab tab tab through completions. Sometimes you get a 100-line diff to your code that nails it, which could have taken 10+ minutes before.
I still don't think I got sufficiently used to all the features. It's a bit like learning to code all over again but I basically can't imagine going back to "unassisted" coding at this point, which was the only possibility just ~3 years ago.
D.Polychronopoulos retweeted
Great to see our research "Genetic factors associated with reasons for clinical trial stoppage" finally out in Nature Genetics
nature.com/articles/s41588-0…
D.Polychronopoulos retweeted
𝐟𝐥𝐞𝐱𝐲𝐧𝐞𝐬𝐢𝐬, our new deep learning-based data integration tool for multi-omics.
1000s of deep learning workflows are available.
Paper: biorxiv.org/content/10.1101/…
Code: github.com/BIMSBbioinfo/flex…
Benchmarks: bimsbstatic.mdc-berlin.de/ak…
Tutorial/docs: bimsbstatic.mdc-berlin.de/ak…
D.Polychronopoulos retweeted
Published today! We found that cancer cells adapting to drug treatments don't simply switch from a sensitive to resistant state; instead there's a ‘resistance continuum’ of resistant phenotypes with epigenetically reprogrammed states. 🧵⬇️
nature.com/articles/s41586-0… @Gustavo_SFranca
D.Polychronopoulos retweeted
What I’m reading this week in single cell omics, generative AI in genomics and medicine, microbial genomics, bioinformatics software, … 🧬🖥️🧵
D.Polychronopoulos retweeted
How to link single-cell diversity and the emergence and maintenance of robust multicellular processes in developed tissues? 🤔
What can we learn about multicellular coordination in tissues from single-cell omics data?
Read about our view: arxiv.org/abs/2403.06753
D.Polychronopoulos retweeted
Opportunities and tradeoffs in single-cell transcriptomic technologies buff.ly/47NcBgv
D.Polychronopoulos retweeted
With six approved drugs, siRNA is now an established therapeutic modality poised for expansion go.nature.com/49tcrf8
rdcu.be/dAsb2
D.Polychronopoulos retweeted
**Is single-cell data integration a solved problem? **
Lots of great methods, but still not quite yet! Specifically, Check Dataset Imbalance!
🎉 Excited to announce that our study on dataset imbalances in single-cell analysis just hit Nature Biotechnology (@NatureBiotech)! 🎉 Dive into our paper titled "Characterizing the impacts of dataset imbalance on single-cell data integration" here: nature.com/articles/s41587-0…
🔬 We've thoroughly assessed how dataset imbalances can twist the results of single-cell integration and batch-correction, affecting everything from cell-type classification to differential expression. Our work doesn't just highlight the issue; it offers solid solutions!
🧩 Starting with balanced PBMC data, we uncovered how changing cell-type balance disrupts integration results. But, when we cluster similar cell-types hierarchically, the magic happens – stability is dramatically improved!
📊 We didn't stop there. Our research extends to how these imbalances influence crucial downstream analyses. The impact? Statistically significant changes that could sway biological conclusions.
🛠️ To empower researchers, we've developed a suite of guidelines and introduced innovative balanced clustering metrics. Our tools are designed to navigate the tricky waters of dataset imbalance, ensuring accurate, meaningful results. Check out our full tutorial: github.com/hsmaan/Iniquitate…
🧬 And, for a cherry on top, we've redefined benchmarking standards with new metrics that ensure fair and nuanced interpretation of integration results. Future benchmarking, we've got you covered! Explore our balanced clustering package: github.com/hsmaan/balanced-c…
🔧 Our comprehensive toolkit, including code, data, and detailed documentation, is now fully open-source! We're setting new standards for method development and benchmarking in single-cell integration, with an eye on the broader horizon of representation learning. Get all the goodies here: github.com/hsmaan/Iniquitate
Huge congratulations to all the authors, especially the first author, Hassaan Maan, a PhD student I co-supervise with @kieranrcampbell, who almost single-handedly led this project!
@UHN @MBPatUofT @VectorInst @UofTCompSci @UofT_LMP @pmcc_ai @UHNAIHUB
D.Polychronopoulos retweeted
Are you looking to apply Large Language Models to biomedical questions? You may want to take a look at our new framework (biochatter.org) and the updated preprint (arxiv.org/abs/2305.06488). 🧵👇
ALT High-level overview of the BioChatter framework and its functionalities
D.Polychronopoulos retweeted
Major Cell2Sentence update 🎉🔬! We’ve been thrilled to see the attention Cell2Sentence has received from the single-cell community.
Now, we’re excited to release our first update of Cell2Sentence (C2S) - a framework to leverage LLMs to train foundational single-cell models, directly in text.
What’s new & out:
Updated preprint with latest results biorxiv.org/content/10.1101/…
First full cell model available on the HuggingFace hub huggingface.co/vandijklab/py…
Updated codebase for data transformation & training github.com/vandijklab/cell2s…
We now fine-tune language models to generate entire cells, predict combinatorial cell labels, and generate textual data insights directly from cell sentences.
We train GPT-2 and Pythia models on a large multi-tissue dataset containing 36M cells from @cellxgene as well as an immune tissue dataset containing 270k cells.
C2S LMs achieve SOTA performance in single-cell data generation.
C2S models trained for combinatorial label prediction settings excel in low-data regimes, outperforming single-cell foundation model baselines.
We also show that C2S models benefit from natural language pre-training and always outperform models trained from scratch on cell sentences.
C2S provides a straightforward approach to adapting LLMs for single-cell data analysis, leveraging their natural language capabilities to generate and derive insights from single cells.
We are convinced that C2S’ approach of integrating data modalities through text is the way forward for single-cell foundation models, from representing multi-omics data to generating clinical insights, all in a human readable format.
We’re excited to start building a community around Cell2Sentence! If you also think that C2S will be the framework for single-cell foundation models, and are interested in contributing, reach out to us! We welcome any collaborations and discussions.
Huge thanks to our collaborator @aminkarbasi and the C2S team (@danielflevine, @sachalevy3, @SyedARizvi5688, @nazreenpm, Xingyu Chen, @dzhang03, @GhadermarziSina, Ruiming Wu, Ivan Vrkic, Anna Zhong, Daphne Raskin, Insu Han, @aho_fonseca, @josueortc) for their hard work on C2S! Special thanks to @rahuldhodapkar, who co-supervises this project.