David Yang

@JoYatesResearch Congrats!

🧵 Introducing SpatialFusion: a lightweight #multimodal #FoundationModel for pathway-informed spatial #niche mapping. 📄 doi.org/10.64898/2026.… It combines #histopathology and #SpatialTranscriptomics to identify functional microenvironments in tissue. Here’s what we found 👇

@HugoAerts Really cool stuff. Congrats!!

Big news: the thymus may be critical for adult health What if we missed something fundamental about the immune system? In two back-to-back papers in Nature, we show thymic health (AI on 30,000+ CT scans) links to longevity, disease risk & immunotherapy outcomes.

@VincentWenxinXu @YuWeiChenMD @DrChoueiri @DanaFarber_GU @MDAndersonnews @Clara__Steiner @lilli_ash @ChadTangMD @DrYukselUrun @yekeduz_emre @PavlosMsaouel @EUplatinum Congrats!

@DrChoueiri @DanaFarber_GU @MDAndersonNews @Clara__Steiner @lilli_ash @ChadTangMD @DrYukselUrun @yekeduz_emre KIM-1 and ctDNA (next-generation, genome-wide assay) are complementary in oligometastatic RCC, and the K-COMPASS model identifies patients with best outcomes after metastasis directed therapy. @ChadTangMD @PavlosMsaouel Check it out in @EUplatinum: sciencedirect.com/science/articl…

The KIM-1 story continues with this joint effort between @DanaFarber_GU @VincentWenxinXu and @MDAndersonNews @ChadTang presented the K-COMPASS model, integrating circulating KIM-1 + ctDNA MRD to risk-stratify oligometastatic ccRCC treated with MDT. - KIM-1 and ctDNA independently associated with systemic therapy-free survival (baseline and 3 mo) - Higher KIM-1 tracked with worse outcomes (baseline PFS HR 2.2 / OS HR 5.1; 3-mo PFS HR 3.5 / OS HR 5.0) - Model performance: C-index 0.76; online tool available (trialdesign.org) Promising integrated blood-based + clinical risk model for risk-adapted decisions. Happy to share that the full publication is now available on @EUplatinum: 10.1016/j.eururo.2026.01.004 #GU26 #RCC @OncoAlert

@laurasenamd @Adam_Weiner535 @CR_AACR @PCF_Science @AACR @hopkinskimmel Congrats!

1/11 Excited to share our study of polyamines in prostate cancer in @CR_AACR! We found that, in models of advanced PCa, supraphysiological androgens (SPA) drive production of polyamines that help the cancer survive. @PCF_Science @AACR @hopkinskimmel aacrjournals.org/cancerres/arti…

@mouwlab Congrats!

Congrats Filipe! An incredible honor and so well-deserved!

@JoYatesResearch @VanAllenLab @val_boeva @DanaFarber @CSatETH Congrats!!

@VanAllenLab @val_boeva @DanaFarber @CSatETH We hope this serves as a foundational resource for the EAC & broader cancer research community.🧬 Full paper here: cell.com/cell-reports-m… (9/9)

Thrilled to share our new paper out today in Cell Reports Medicine! @VanAllenLab @val_boeva We map cell states & neighborhoods across clinical stages of esophageal adenocarcinoma (EAC) using single-cell, epigenomic & spatial data.🔗cell.com/cell-reports-m… (1/9)

@sjohri20 @NatureComms @VanAllenLab Congrats!!

🚀Excited to share our new paper, now published in @NatureComms! ✨ We present BEANIE, a nonparametric method that reduces false positives when analyzing differential gene signature expression in multi-patient clinical scRNA-seq cohorts @VanAllenLab 🧪🧵nature.com/articles/s4146…

@SashaGusevPosts Congrats!!

Thrilled to receive the Presidential Early Career Award for Scientists and Engineers (PECASE)! Could not have done this without my amazing lab members and mentors. whitehouse.gov/ostp/news-upda…

@PCFnews @urotoday Thanks for giving @MtkingMD and me the opportunity to speak about our work!

Researchers Drs. Martin King & @DYangMD discuss AI's potential to automate and standardize prostate tumor assessment. Could this revolutionize clinical decision-making in #oncology? Watch the full discussion on @urotoday - bit.ly/4fNC9O7

@DKang_PhD @fredhutch Congrats!!

Off to a new start at @FredHutch as Assistant Professor - I can’t be grateful enough to my incredible mentors/colleagues/friends at DFCI for their support. It’s hard to leave them behind, but I’m looking forward to exploring new opportunities at FH! #ExerciseOncology

Tonight we sit with the cherished memories of a truly unforgettable individual who changed the lives of so many fighting cancer. Tomorrow we stand and get back to the fight against the terrible disease that took one of our best, brightest, and kindest. Love you forever, Felix.

@DongNguyeb @KatieLeeMDPhD @MtkingMD 2. The AUC for the RP cohort was 0.89 vs 0.79 (p=0.25, though likely underpowered due to fewer events). We chose 7y for RT and 5y for RP to be close to the median f/u of both cohorts.

Response to your tweetorial. 1/ The F1-score is high training/testing cohort, but in what cut-off? The cut-off that they choose is the dice coefficient >10%, this is very low cut-off that show the imaging detected by AI is poorly fit to reference. 2/Vai is greater discrimination AUC than NCCN but it is just only one time point (7 year) and only in one cohort (RT groups). These hidden is the whole time follow up and the other cohort (surgery cohort)

Excited to share work with coauthors Leslie Lee, @KatieLeeMDPhD @MtkingMD and others on "AI-derived Tumor Volume from Multiparametric MRI and Outcomes in Localized Prostate Cancer." A tweetorial: 1/n

@DongNguyeb @KatieLeeMDPhD @MtkingMD Thanks for your interest. 1. Yes, we chose Dice >=10%, which is the cutoff used by the PI-CAI grand challenge, to allow for standardization of comparisons.

@KatieLeeMDPhD @MtkingMD Many thanks to the coauthors and @PCFnews for the support 8/n

@KatieLeeMDPhD @MtkingMD If further validated, AI-determined tumor volume may become a novel biomarker for improving risk stratification for pts w/ localized PCa (in addition to genomics and computational pathology approaches) 7/n

@KatieLeeMDPhD @MtkingMD The AI-determined tumor volume may also have greater performance for predicting MFS than NCCN risk groups (e.g. AUROC for 7y MFS 0.84 vs 0.74, p=0.02 for RT-treated patients) 6/n

@KatieLeeMDPhD @MtkingMD Particularly interesting was our observation the intraprostatic tumor size was associated with both BCR and MFS for both RT and RP-treated patients, even after adjusting for the clinical, radiologic, and pathologic characteristics 5/n

@KatieLeeMDPhD @MtkingMD Using a cohort of 732 patients w/ PCa and mpMRIs, we built a segmentation model with the nnU-Net approach which showed good performance (e.g. F1 scores of 84-87% for PI-RADS 3-5 lesions across training/test cohorts) 4/n

@KatieLeeMDPhD @MtkingMD We envisioned that this model would be useful in several ways, such as for focal microboosts for EBRT: redjournal.org/article/S0360-… 3/n

@KatieLeeMDPhD @MtkingMD We initially became interested in building a deep learning model for segmenting the intraprostatic tumor from prostate mpMRIs (though noting that we were certainly not the first to attempt this task, e.g. thelancet.com/journals/lanon…) 2/n