Gamze Gürsoy

Big congratulations to @luisasantus.bsky.social for winning one of the awards for good open data management in her research. 👏🎊

Luisa is currently visiting @gamzeandgursoy.bsky.social lab at @columbiauniversity.bsky.social during her PhD in Cedric Notredame's Lab at @crg.eu
#OAWeek2025

Excited to share that our paper that was accepted to #CHIL2025 "KEEP: Integrating Medical Ontologies with Clinical Data for
Robust Code Embeddings" is now available at PMLR: raw.githubusercontent.com/mlresearch/v...

Many thanks to the MacMillan Center for the Study of the Non-coding Cancer Genome @nygenome for making this work possible, and @ColumbiaDBMI for always being so supportive of our work. (4/4)

We think this approach can have a lot of potential in downstream tasks such as high-resolution enhancer-gene linkages, hypothesis generation, annotating non-coding variants, and even perhaps for fine mapping (3/n)

We developed a physical modeling based approach to "fill in the blanks" in the Hi-C matrices, which allows unsupervised imputation of high-resolution matrices without needing a high-resolution training dataset. (2/n)

pC-SAC: A method for high-resolution 3D genome reconstruction from low-resolution Hi-C data Abstract. The three-dimensional (3D) organization of the genome is crucial for gene regulation, with disruptions linked to various diseases. High-throughpu

Super excited to share the first 3D genome paper from my lab @narjournal.bsky.social , led by my smart and hardworking student Carlos (joint with @LuLaboratory) with assists from a brilliant summer intern Narjis. (1/n) academic.oup.com/nar/article/...

that is homomorphic encryption friendly. It's complexity is independent of the number of reads. We think the optimizations in this algorithm can be used in many plaintext NGS processing algorithms too! 2/2

Secure and scalable gene expression quantification with pQuant - Nature Communications RNA-seq data can expose private genotypes during computation. Here, the authors present pQuant, a privacy-preserving tool that uses homomorphic encryption to quantify gene expression securely, achievi...

super excited to share the newest from my lab: a tool that can securely process RNA-seq reads to quantify gene expression. Huge congrats to the dream team Seungwan, @BioConorWalker, and Annie! We developed a new gene expression quantification algorithm 1/2 www.nature.com/articles/s41...

Stop collaborate and listen! 😀Yesterday was a great day for science! Thank you @standupforscience.bsky.social !The real work begins now. People are listening! We need to challenge misunderstandings and show that investment in science is crucial for our health for our economy and for our future!

One of the most powerful moments of the day came from Emily Whitehead who told the story of how she was the first pediatric patient to receive CAR T-cell therapy for her leukemia at age 5: “I stand up for science because science saved my life. And that’s a fact.” @standupforscience.bsky.social

Comprehensive dissection of cis-regulatory elements in a 2.8 Mb topologically associated domain in six human cancers - Nature Communications The oncogene MYC plays a key role in cancer initiation and progression. Using thousands of CRISPR perturbations, the authors investigate regulators of MYC in six different cancers. These tumor-specifi...

🔷 📰 BLUETORIAL time! 📰 🔷

Excited to share recent work from our lab on noncoding genomics 🔎🧬: High-resolution CRISPR perturbations of the MYC TAD (~3 Mb) in 6 different human cancer cell lines
www.nature.com/articles/s41...

This metric allows us to determine (in a principled way) which NN layers would benefit from federation and which layers should be locally trained when data is non-IID. Please let us know if you have a study that can benefit from implementing this tool! (2/2)

PLayer-FL: A Principled Approach to Personalized Layer-wise Cross-Silo Federated Learning Non-identically distributed data is a major challenge in Federated Learning (FL). Personalized FL tackles this by balancing local model adaptation with global model consistency. One variant, partial F...

More science! We have a new pre-print - more on non-IID federated learning: We (well my student Ahmed) were inspired by model pruning and developed a new metric called federation sensitivity. (1/2) www.arxiv.org/abs/2502.08829

Many thanks to @columbiauniversity.bsky.social and @nygenome.org - I cannot emphasize more that without resources, facilities, and admin support, none of the science is possible! (5/n)

and many many thanks to @tuuliel.bsky.social for helping us greatly with the deep dive to our results, robustness, as well as making the paper readable for the audience we targeted. We are extremely lucky to be surrounded by great scientists from many disciplines! (4/n)

This work wouldn't be possible without help from our great collaborators. Peihan helped tremendously to understand the theories behind MPC and its implementation (3/n)

She developed clever algorithms to make eQTL mapping Multiparty Computation friendly and performed a very rigorous analysis of its accuracy, performance, and security. Many thanks to our visiting student Yebin for the assist as well. (2/n)

Secure and federated quantitative trait loci mapping with privateQTL Choi et al. developed a novel tool for privacy-preserving cross-institutional eQTL mapping studies. The authors benchmarked their tool against meta-analysis and demonstrated that it achieves higher ac...

The world is on fire but we must continue doing science! With immense pride, I share the latest from my lab: privateQTL - a method for federated and secure eQTL mapping, led by my brilliant student Annie from @ColumbiaDBMI www.cell.com/cell-genomic...
(1/n)

Call for submission to 2025 @recombconf.bsky.social Satellite Conference on Biomedical Privacy and Equity (RECOMB P&E) is open! For more info please see recomb-privacy-equity.github.io/papers/ or reach out to our wonderful chairs (Hoon Cho and Shalmali Joshi! )

🎇🎄Some holiday reading: Overdue TWEETORIAL on the latest preprint from our lab 🎇🎄

In this work, superstar postdoc @xinhexue.bsky.social combined 2 kinds of pooled CRISPR screens to pinpoint noncoding regulatory elements and the transcription factors that activate these elements.

Many thanks to the lead authors for bringing this idea to a different level and making this tool to scale to biobanks! We dedicate this paper to the late Dr. Lehner, whose visionary ideas inspired this work. His presence is deeply missed!

SQUiD: ultra-secure storage and analysis of genetic data for the advancement of precision medicine - Genome Biology Cloud computing allows storing the ever-growing genotype-phenotype datasets crucial for precision medicine. Due to the sensitive nature of this data and varied laws and regulations, additional securit...

New paper alert! 🚨 SQUiD: Our secure and queryable genotype-phenotype database prototype with many secure analysis functionalities is out
@genomebiologyunit.bsky.social link.springer.com/article/10.1...

Super excited for my lab's first paper at @recombconf.bsky.social. See you all in South Korea!