In partnership with
Tech moves fast, but you're still playing catch-up?
That's exactly why 200K+ engineers working at Google, Meta, and Apple read The Code twice a week.
Here's what you get:
Curated tech news that shapes your career - Filtered from thousands of sources so you know what's coming 6 months early.
Practical resources you can use immediately - Real tutorials and tools that solve actual engineering problems.
Research papers and insights decoded - We break down complex tech so you understand what matters.
All delivered twice a week in just 2 short emails.
Hello {{First Name|there}},
In metastatic clear cell renal cell carcinoma (ccRCC), angiogenesis is both a biological feature and a primary therapeutic target. Yet despite the widespread use of VEGF-tergeted therapies, clinical responses remain highly variable
A critical question persists:
Can we identify which tumors are truly angiogenesis-dependent, and therefore most likely to respond to anti-angiogenic therapy?
A 2024 study introduces a visually interpretable deep learning biomarker derived directly from routine H&E histopathology that may offer a clinically scalable answer.
From RNA Signatures to Digital Histopathology
Prior analyses from the IMmotion150, IMmotion151, and JAVELIN Renal 101 trials demonstrated that tumors with high expression of a 6-gene angiogenesis signature (VEGFA, KDR, ESM1, PECAM1, ANGPTL4, CD34), often referred to as the RNA Angioscore, correlates with improved progression-free survival on sunitinib.
These findings are now being prospectively evaluated in the OPTIC RCC trial.
However, transcriptomic assays face real-world barriers:
Cost and turnaround time
Batch variability
Limited sampling of highly heterogeneous tumors
Feasibility constraints in small biopsies
The question posed by this new study:
Can angiogenesis be quantified directly from standard H&E slides, without RNA sequencing?
A Deep Learning Angioscore from H&E Alone
To address this, investigators developed an interpretable deep learning (DL) model trained on:
RNA-based Angioscores (from TCGA and trial datasets)
CD31 immunohistochemistry (IHC) to define endothelial “ground truth”
Unlike traditional “black box” models, this approach generates pixel-level vascular mask from H&E images.
The final H&E DL Angioscore is calculated as the fraction of pixels predicted to represent endothelial structures, providing a direct, visual representation of tumor vasculature.
This dual-modality training (RNA + CD31) enhances both clinical transparency and interpretability.
Validation and Correlations Across Cohorts
The new H&E DL Angioscore was evaluated across multiple datasets.
The H&E-based score successfully surpassed the gold standard RNA angioscore and treatment response. Notably, the inclusion of the CD31 IHP did not enhance the performance of the H&E DL angioscore compared to the gold standard. They note that additional studies are needed to determine if this is due to difficulties in quantifying CD31 IHC, or whether the combined training leads to significant differences in vasculature.
Beyond prediction of RNA expression, biological correlations were demonstrated:
Inverse correlation with:
WHO/ISUP nuclear grade
Advanced TNM stage
Tumor size
Sarcomatoid features
This supports an emerging paradigm in ccRCC:
As tumors become more aggressive, vascularity decreases.
Genotype associations were also observed:
BAP1 loss → significantly lower angiogenesis
PBRM1 loss → modest, non-significant increase
Why this Matters
Angiogenesis is a hallmark of cancer, and ccRCC is its archetypal example.
This study highlights that quantitative vascular phenotyping may be extractable directly from standard histopathology. This could transform achieved H&E slides into a scalable, low-cost biomarker platform.
Future directions include:
mapping intra-tumor angiogenic heterogeneity
Integrating immune and vascular microenvironment features
Combining histopathology with radiology and genomics
developing fully image based predictors
If validated prospectively, this approach could reduce reliance on costly molecular assays and move precision oncology closer to routine diagnostic workflows.
As the authors write “The H&E DL Angioscore enables prediction of anti-angiogenic therapy response in ccRCC solely from H&E images, offering a cost-effective and interpretable alternative to RNA-based assays. By bridging the gap between molecular insights and clinical feasibility, our work sets the stage for transformative advancements in ccRCC therapeutics.”
Check out this weeks youtube video to see our mascot Dr. Angio bringing complex health and research topics to life.
Check out the Angiogenesis Foundation Green Tea from Harney & Sons to support your vascular health.
Best wishes,
- The Angiogenesis Foundation
P.S. Like what you’re reading? Support our mission to advance research and share science-backed health insights.
Make a donation to the Angiogenesis Foundation today.
Learn how to code faster with AI in 5 mins a day
You're spending 40 hours a week writing code that AI could do in 10.
While you're grinding through pull requests, 200k+ engineers at OpenAI, Google & Meta are using AI to ship faster.
How?
The Code newsletter teaches them exactly which AI tools to use and how to use them.
Here's what you get:
AI coding techniques used by top engineers at top companies in just 5 mins a day
Tools and workflows that cut your coding time in half
Tech insights that keep you 6 months ahead
Sign up and get access to the Ultimate Claude code guide to ship 5X faster.
