✨New paper alert @CD_AACR✨ .- GDD-ENS distinguishes 38 cancer types (>96% of patients) w/ high accuracy using routinely collected DNA sequencing data #MSKIMPACT and generalizes well to other gene panels w/o retraining.- CNAs (arm/gene level) are key features for classification.

RT @adamcweiner: Happy to announce that PERT is now published in @NatureComms ! With PERT, we explore replication timing variability (Fig 3….

RT @MSKEducation: Meet #MSKPostdoc Cassandra Burdziak (@CassBurdziak) from the @dana_peer Lab, who uses single-cell technologies and algori….

RT @quaidmorris: Metient reconstructs the migration histories of metastatic cancers by scoring a Pareto front of parsimonious histories. It….

RT @CD_AACR: New from the June issue—.#DeepLearning Model for Tumor Type Prediction using Targeted Clinical Genomic Sequencing Data, by @mm….

RT @DoronTheViking: Super excited to announce that our work “Wasserstein Wormhole: Scalable Optimal Transport with Transformer” has been ac….

RT @CornellinDC: Graduate education is critical for the future. @Cornell @CornellGrad @cornell_tech @WeillCornellGS @WeillCornell grad stud….

RT @DoronTheViking: Woohoo ENVI/COVET is out in @NatureBiotech!! Couldn’t have done it without my brilliant mentor @dana_peer and the rest….

RT @OnlyInBOS: A statue has been unveiled on Mile 2 of the Boston Marathon in honor of Spencer, the official dog of the Boston Marathon, wh….

RT @CD_AACR: #OnlineFirst: Deep Learning Model for Tumor Type Prediction using Targeted Clinical Genomic Sequencing Data, by @mmdarmofal, @….

Thanks to an incredible team for making this work possible! @MFBerger1 @quaidmorris @JieFuChen2 @evakiani @AnnaVarghese4 @nikolausschultz @cbmstudent + x-less Shalabh, Gurnit, Michael, Jason, Anoop, Aijaz, & the GDD-RF authors (.

Here’s a link to the summary of the pre-print ( but please check out our manuscript, because there have been a lot of improvements due to an extremely useful and smooth review process. Thanks @ElizSMcKenna and our reviewers!.

RT @MildredUnti: So excited to announce that my most recent work has been published. If you’re interested in genetically encoding circular….

RT @NobelPrize: “Ten years ago I was kicked out and forced to retire.”. Our new medicine laureate Katalin Karikó (@kkariko) told us how mu….

We would also like to especially acknowledge Alex Penson, Niedzica Camacho, @zheng_youyun and @BarryTaylorLab whose initial GDD-RF model was the foundation for this work, without which our updated model would not be possible.

Thank you to @quaidmorris, @MFBerger1, Shalabh Suman, Gurnit Atwal, Jason Chang, @JieFuChen2 , Mike Toomey, @evakiani, @AnnaVarghese4, Anoop Balakrishnan Rema, Aijazuddin Syed and @nikolausschultz for making this possible! @cbm_student @MSKCancerCenter #MSKIMPACT.

GDD-ENS has already been implemented prospectively, concurrent with MSK-IMPACT clinical testing. It provides prediction confidence and prediction-specific feature importance values to clinicians on an internal database 🧬🖥️.

Even GDD-ENS misclassifications are informative. Most reflect well-known tumor biology, but we also found that non-uveal melanomas misclassified as uveal are from the rare CNS melanoma class that GDD-ENS Shapley values reveal as having uveal-like pathway alterations.

GDD-ENS supports nearly a 4x increase in # of patients with cancers of unknown primary (CUP) that can be connected to a FDA-approved targeted therapy, up to 48% of patients with actionable alterations.

Additional, patient-specific clinical features can be incorporated into GDD-ENS without retraining. Adding biopsy site and histology for metastases increases the overall accuracy by 4% and the proportion of predictions that are high confidence from 72% to 80%