Excited to share Dr. Elkin's NetFlow manuscript — a framework for constructing interpretable graph representations of high-dimensional biomedical data. It captures both clustering and continuous variation to reveal biologically meaningful structure. #AI #Oncology #Bioinformatics
(1/2) Lung cancer can co-opt genes that normally help a fetus develop and evade the mother’s immune system, according to research from Dr. Jung Hun Oh.
And while these pregnancy-specific glycoproteins (PSGs) can get activated in the cancers of both men & women, female patients had poorer outcomes.
9/ Want to try ORCO? If you're working with omics data & want to see how network curvature can enhance your analysis, I’d love to help! Please reach out.
#Bioinformatics #NetworkBiology #SystemsBiology
8/ ORCO is open-source & [relatively] easy to use! 🎉 Install via pip and start exploring network robustness in your data. Please reach out if you find any bugs or issues!
📌 Code: github.com/aksimhal/ORC...
7/ Past successes of ORC:
🔹 Identified novel gene signatures for high-risk multiple myeloma (Simhal et al. 2023)
🔹 Revealed therapeutic targets in sarcoma (Elkin et al. 2024)
🔹 Improved graph neural networks for cancer survival prediction (Zhu et al. 2023)
6/ ORCO has already provided insights into multiple cancers 🦠 and neurodevelopmental disorders 🧠 by highlighting network vulnerabilities and functional cooperation in gene signaling.
5/ How does ORCO work?
✅ Input: a biological network (e.g., gene or protein interactions)
✅ Input: omics data (RNA-seq, proteomics)
✅ Output: Edge-based values that describes the robustness between nodes (e.g., genes).
4/ Why does this matter? ORCO focuses on interactions—revealing how biological systems maintain function under stress (or break down when fragile). This can uncover new disease mechanisms & drug targets!
3/ How does ORC work?
🔹 At a very high level, if two genes have many connections, their interaction has positive curvature = robust.
🔹 If a connection is a single weak link, easily disrupted, it has negative curvature = fragile
2/ ORCO is a network analysis tool that applies Ollivier-Ricci curvature (ORC) to omics data. It identifies robust and fragile network interactions, helping uncover key patterns of dysregulation. Let's break it down! 🧵
🚀 Just published in Bioinformatics! Introducing ORCO: Ollivier-Ricci Curvature-Omics, a python package for analyzing robustness in biological systems. A 🧵 on what it is, why it matters, and how you can use it. With @joedeasy.bsky.social and the great team at @mskcancercenter.bsky.social
ORCO: Ollivier-Ricci Curvature-Omics: an unsupervised method for analyzing robustness in biological systems https://www.biorxiv.org/content/10.1101/2024.10.06.616915v1
13/ Thanks to @vincentrk.bsky.social & the Blood Cancer Journal team for bringing this research out!
12/ Bottom line:
🔹 High WEE1 = worse survival in MM
🔹 WEE1 is independent of known risk factors
🔹 Targeting WEE1 might be a new therapeutic avenue
🔗 Read more: www.nature.com/articles/s41...
11/ Machine learning insights 🤖
Random survival forest showed that WEE1 expression alone has as much prognostic power as ISS staging.
Random forest modeling of the local WEE1 genomic network showed overexpression of WEE1is independent of any 1-hop network genes.
10/ More work to do to figure out the TP53 WEE1 connection, but differential gene expression analysis implicated the hallmark P53 pathway. Faulty P53 function may lead to a larger reliance on WEE1.
9/ The P53 connection 🧬
Differences in PFS among patients with TP53 deletions when stratifying by WEE1 expression were massive! Our results seemed to show that patients with TP53 deletions but without high WEE1 expression may not be at high risk after all.
8/ So why is this important?
✅ WEE1 expression could be used as a new prognostic biomarker in MM.
✅ WEE1 inhibitors, already in trials for other cancers (including those by Zentalis & Debiopharm) might be a therapeutic option for MM patients.
7/ We then validated our findings using GEP datasets from the Total Therapy 2 and3 trials. The same pattern emerged — WEE1 was a consistent predictor of poor survival.
6/ Interestingly, high WEE1 expression predicted worse PFS independently of other high-risk factors like MYC translocations, chromothripsis, TP53 deletion, and independent of treatment used!
5/ This observation was seen only on the RNA level! There was no genetic signature predictive of high WEE1 expression, providing a rationale for integrating GEP data into outcome prediction instead of only looking at genomics.
4/ Using @themmrf.bsky.social CoMMpass dataset, we identified high-risk and low-risk groups based on WEE1 expression. Results? High WEE1 expression = significantly worse progression-free survival.
3/ WEE1 is a tyrosine kinase that regulates the cell cycle and plays a key role in DNA damage repair and tumor progression. Abnormal WEE1 expression has been implicated in breast, ovarian, and gastric cancers, but its role in MM has been largely unexplored!
2/ Current MM prognostic tools rely on disease burden & a limited set of genomic markers. We identified WEE1 as a prognostic marker of interest through an Ollivier-Ricci curvature based network analysis. We analyzed MM patient data & found WEE1 to be a strong predictor of survival.
1/ 🚨New research alert! 🚨Our study in Blood Cancer shows that high WEE1 expression is an independent predictor of poor survival in multiple myeloma, with @joedeasy.bsky.social, @malinhultcrantz.bsky.social, @szusmani.bsky.social, and the rest of the @mskcancercenter.bsky.social team! 🧵👇#MyelomaSky
