Andrea Schmidts

Nanocarrier imaging at single-cell resolution across entire mouse bodies with deep learning - Nature Biotechnology An integrated experimental and deep learning pipeline reveals cell-level targeting of nanocarriers in whole bodies.

New paper from our lab (@erturklab.bsky.social)

We show how tissue clearing, light sheet microscopy, and AI-based image analysis can be combined to determine the biodistribution and drug delivery efficiency of nanocarriers throughout entire mouse bodies with single-cell resolution.

We also want to highlight a very exciting study that was recently published showing efficient DNB CAR-T in the context of glioblastoma (PMID 39420062). (10)

We’re grateful to our funding sources, specifically @dfgpublic.bsky.social, and to our mentors Marcela Maus and Keith Joung. (9)

Thanks a lot to our fantastic team of the Schmidts & Grünewald Labs and all collaborators. Special thanks to our exceptionally talented MD and PhD students, Markus Mergen and Daniela Abele, who led this study. We’re very excited about the upcoming experiments! (8)

This is just a first small step for #AI in #PrecisionMedicine. We’re working on many more AI-guided CAR-T cell approaches and hope that #GenerativeAI can democratize and greatly accelerate binder design for new #celltherapies and #genetherapies, especially in the context of cancer resistance. (7)

The bispecific antibody teclistamab was recently studied for resistance, with variant R27P mediating antigen escape (PMID 37653344). Resistance emerges after 9 months in most patients. Our binder B10, with a different predicted binding interface, enabled #aiCART cells to kill BCMA-R27P cells. (6)

Given the high rates of resistance to immunotherapies and CAR-T (>50% resistance to Abecma and Carvikty after 2 years), we wondered if our new candidate could bind mutated variants of BCMA. (5)

Next, we used the #GenerativeAI tool RFdiffusion (PMID 37433327) from the Baker lab to design DNBs against BCMA, a key target in CAR-T therapy. After in vitro characterization in the CAR-T context, we selected the lead candidate (B10) and demonstrated in vivo activity in a #myeloma mouse model. (4)

We initially tested an in silico generated de novo binder (DNB) from the Baker Lab targeting EGFR (PMID 35332283). In a CAR construct, this binder performed well- killing cancer cell lines and patient-derived pancreatic cancer #organoids & forming immunological synapses with target cells. (3)

In the first preprint from my and @grunewald.bsky.social labs @TU_Muenchen, we show how this could work for new #celltherapies and #genetherapies. (2)

Cancer escapes immunotherapies by mutating its surface, meaning even groundbreaking #CARTCell therapies fail too often. But what if #GenerativeAI could design CAR-T with de novo binders faster, targeting resistant tumors? 🧵 www.biorxiv.org/content/10.1... (1)