Lightnews — Scholar-powered news

Kieran Campbell

@kieranrcampbell.bsky.social

3.9K followers 490 following 120 posts

https://www.camlab.ca | ML for cancer genomics + single-cell and spatial computational methods | Toronto Canada 🖥️🧬

www.camlab.ca

Posts Media Videos Starter Packs

Pinned

Kieran Campbell @kieranrcampbell.bsky.social · Jan 6

How can we quantify single cell expression from multiplexed imaging datasets given segmentation errors?

To help tackle this, we introduce evaluation metrics, a gold-standard dataset, and STARLING—a new probabilistic machine learning method (1/4)

www.nature.com/articles/s41...

Segmentation aware probabilistic phenotyping of single-cell spatial protein expression data - Nature Communications

Spatial expression assays are affected by segmentation errors leading to difficulty interpreting cell types. Here, the authors introduce a machine learning model to infer cell types accounting for suc...

www.nature.com

2 10 29

Kieran Campbell @kieranrcampbell.bsky.social · 14d

Congratulations!

1 1

Reposted by Kieran Campbell

Research at Sinai Health @sinaihealth.bsky.social · 15d

⏰Reminder: There's still time to apply!

LTRI is hiring two outstanding Principal Investigators in Population Health Data Science.

Find out more and apply by October 20th, 2025 ➡️ apply.interfolio.com/172529

4 1

Kieran Campbell @kieranrcampbell.bsky.social · 28d

Congrats Trevor very exciting!

Reposted by Kieran Campbell

Trevor Graham @trevorgraham.bsky.social · 28d

Studying cancer evolution needs multi-region or single cell seq for phylogenetics, right? Amazingly (I think!) we found single-sample bulk methylation suffices, via analysis of "fluctuating methylation". In @nature.com today led by brilliant @calumgabbutt.bsky.social www.nature.com/articles/s41...

Fluctuating DNA methylation tracks cancer evolution at clinical scale - Nature

Cancer evolutionary dynamics are quantitatively inferred using a method, EVOFLUx, applied to fluctuating DNA methylation.

www.nature.com

7 38 91

Reposted by Kieran Campbell

Oded Rechavi @odedrechavi.bsky.social · 28d

Full professor complaining to an assistant professor on how hard it is to get funding nowadays

2 26 210

Kieran Campbell @kieranrcampbell.bsky.social · Sep 3

Fantastic hosting @alexr.bsky.social in Toronto today discussing the hype of AI in cancer biology and a realistic route to leverage it for personalized immunotherapy

2 6

Reposted by Kieran Campbell

Richard White @whitefishlab.bsky.social · Aug 28

Cancer cells are notoriously plastic, but what triggers the changes in cell state? In our new Nature (@nature.com)‬ paper, Miranda Hunter (@mrndhntr.bsky.social‬) found that mechanical pressure is a key to this switch. Check it out here: go.nature.com/47iwDm2

7 21 59

Reposted by Kieran Campbell

Daniel Durocher @durocher1.bsky.social · Aug 28

JOB ALERT 🚨 We are hiring TWO principal investigators in cell, molecular, systems, or chemical biology in Toronto, Canada at @sinaihealth.bsky.social. We provide a generous startup, fully funded salary and academic appointment at U of Toronto.

www.nature.com/naturecareer...

Please repost!

170 130

Reposted by Kieran Campbell

Mike-ghoul Half-man @michaelhoffman.bsky.social · Aug 18

"The perfect segmentation doesn't ex—"

—@kieranrcampbell.bsky.social. www.nature.com/articles/s41...

A Segmentation errors in highly multiplexed imaging experiments induce clusters formed by the composition of multiple true cell types. B Our probabilistic model STARLING for clustering highly multiplexed imaging data models whether each cell is observed segmentation error free, and if not models the composition of cell types leading to the observed cell. The loss function jointly maximies the likelihood of the data and the ability to discriminate on-the-fly simulated segmentation errors. C STARLING takes as input summarized single-cell expression and cell sizes as output by common pipelines for multiplexed imaging, along with a cluster initialization either user-defined or derived from multiple existing algorithms. D The output includes both the denoised cell types along with a per-cell estimate of segmentation errors.

1 2 24

Kieran Campbell @kieranrcampbell.bsky.social · Aug 13

Hopefully there was free pizza!

1 2

Reposted by Kieran Campbell

Adam MacLean @adamlmaclean.bsky.social · Aug 12

Online now! Tumor dynamics & spatial proteomics-informed personalized predictions of growth & response

Beyond RECIST: mathematical modeling and Bayesian inference reveal the importance of immune parameters in metastatic breast cancer

www.biorxiv.org/content/10.1...

5 12

Reposted by Kieran Campbell

Constantin Ahlmann-Eltze @const-ae.bsky.social · Aug 4

Our paper benchmarking foundation models for perturbation effect prediction is finally published 🎉🥳🎉

www.nature.com/articles/s41...

We show that none of the available* models outperform simple linear baselines. Since the original preprint, we added more methods, metrics, and prettier figures!

🧵

Beeswarm plot of the prediction error across different methods of double perturbations showing that all methods (scGPT, scFoundation, UCE, scBERT, Geneformer, GEARS, and CPA) perform worse than the additive baseline.

Line plot of the true positive rate against the false discovery proportion showing that none of the methods is better at finding non additive interactions than simply predicting no change.

2 57 130

Reposted by Kieran Campbell

Wolfgang Huber @wkhuber.bsky.social · Jul 29

The grand vision of the AI university. Professors design curricula, lecture notes and exam questions using genAI. Students read the notes and submit essays and exam answers using LLMs. Massive efficiency and speed gains ensue. Students can go play pool, professors write grants in the saved time.

2 2 21

Reposted by Kieran Campbell

Hannah Meyer @hannahvmeyer.bsky.social · Jul 25

Check out how we tackled one of the fundamental challenges in immunology, out today @cp-cellsystems.bsky.social: www.cell.com/cell-systems.... Our major innovations are BATCAVE and BATMAN, two new players to get us closer to solving TCR-pMHC predictions!

T cell receptor cross-reactivity prediction improved by a comprehensive mutational scan database

Banerjee et al. present a comprehensive T cell receptor cross-reactivity database and a hierarchical Bayesian framework to predict T cell receptor targets. Together, they reveal structural and biochem...

www.cell.com

3 11 19

Reposted by Kieran Campbell

Mike Hallett @hallettmiket.bsky.social · Jul 8

The DNA of pure DCIS. We look at prognosis (invasive & in situ local recurrence) & benefit to radiotherapy in a unique cohort of Ontarians. A manuscript with my long term collaborators Eileen Rakovitch (Sunnybrook) & @vdumeaux.bsky.social breast-cancer-research.biomedcentral.com/articles/10....

Mutational landscape of pure ductal carcinoma in situ and associations with disease prognosis and response to radiotherapy - Breast Cancer Research

Background Managing Ductal Carcinoma in Situ (DCIS) remains challenging due to the lack of reliable biomarkers to predict radiotherapy (RT) response, leading to both overtreatment of indolent disease ...

breast-cancer-research.biomedcentral.com

3 5

Reposted by Kieran Campbell

danapeer.bsky.social @danapeer.bsky.social · Jul 3

I am thrilled to share with the world a new preprint, spearheaded by the truly amazing Jose Reyes in partnership with the Lowe lab. The paper sheds light on benign to malignant transition in PDAC, uncovering roles of plasticity, tissue remodeling and P53. -> www.biorxiv.org/content/10.1...

Oncogenic and tumor-suppressive forces converge on a progenitor-orchestrated niche to shape early tumorigenesis

The transition from benign to malignant growth is a pivotal yet poorly understood step in cancer progression that marks the shift from a pathologically inert condition to a clinically lethal disease. ...

www.biorxiv.org

1 7 41

Reposted by Kieran Campbell

Anshul Kundaje @anshulkundaje.bsky.social · Jun 19

@saramostafavi.bsky.social (@Genentech) & I (@Stanford) r excited to announce co-advised postdoc positions for candidates with deep expertise in ML for bio (especially sequence to function models, causal perturbational models & single cell models). See details below. Pls RT 1/

1 40 55

Kieran Campbell @kieranrcampbell.bsky.social · Jun 16

Six days left to apply!

Kieran Campbell @kieranrcampbell.bsky.social · May 27

🧬 We're hiring - join our lab in Toronto, Canada! 🇨🇦

Exciting postdoc joint with @hartjackson.bsky.social developing machine learning methods for the next generation of spatial histopathology technologies joint with industry partners in the Bay Area.

👉 apply.interfolio.com/168307

Apply - Interfolio {{$ctrl.$state.data.pageTitle}} - Apply - Interfolio

apply.interfolio.com

4 1

Kieran Campbell @kieranrcampbell.bsky.social · Jun 3

As a teenager I went up to Morrissey outside his concert venue in my hometown and asked for an autograph. He insisted he was a lookalike. To this day I don’t know whether it was or not

1 3

Reposted by Kieran Campbell

Quaid Morris @quaidmorris.bsky.social · Jun 3

Please check out our new approach to modeling somatic mutation signatures.

DAMUTA has independent Damage and Misrepair signatures whose activities are more interpretable and more predictive of DNA repair defects, than COSMIC SBS signatures 🧬🖥️🧪

www.biorxiv.org/content/10.1...

Damage and Misrepair Signatures: Compact Representations of Pan-cancer Mutational Processes

Mutational signatures of single-base substitutions (SBSs) characterize somatic mutation processes which contribute to cancer development and progression. However, current mutational signatures do not ...

www.biorxiv.org

17 41

Reposted by Kieran Campbell

Victor Javier @vjsanchez.bsky.social · May 27

Spatial Biology and Organoid Technologies Reveal a Potential Therapy-Resistant Cancer Stem Cell Population in Pancreatic Ductal Adenocarcinoma 🧪 www.biorxiv.org/content/10.1...