SBME Symposium 2026

Beyond Bench to Bedside: Why Biomedical Translation Requires Community, Power, and Partnership

Translational research in biomedical engineering is often framed as a linear pathway from bench to bedside. Yet this model overlooks a critical dimension: the communities who are ultimately meant to benefit from scientific innovation.

Without meaningful engagement, equitable study design, and attention to power, translational efforts risk reproducing the very inequities they aim to address.

Drawing from my work in Indigenous genomics and bioethics, this talk argues that translation does not occur without community. I will explore how positionality, lived experience, and cultural context shape not only how research is conducted, but which questions are asked, how studies are designed, and how outcomes are interpreted.

I will highlight Indigenous approaches to data governance and community-engaged research as models for rethinking biomedical innovation—where participants are partners, not subjects, and where ethical responsibility extends beyond compliance to accountability and relationship.

Ultimately, this talk challenges the biomedical engineering community to move beyond narrow definitions of translation and to consider how integrating community from the outset can lead to more just, effective, and meaningful scientific outcomes.

Dr. Krystal Tsosie’s Biography

Dr. Krystal Tsosie (Diné/Navajo Nation), PhD, MPH, MA, is an Indigenous geneticist-bioethicist and Assistant Professor in the School of Life Sciences at Arizona State University. She is also a genomic data systems architect whose work focuses on building the infrastructure that shapes the future of precision medicine and genomics.

Trained in genetic epidemiology and applied ethics, her research spans human genomics, paleogenomics, and biodiversity genomics, integrating bioethics and data science to develop scalable, community-engaged data systems. She leads the development of Indigenous-led platforms—including biobanks and tribal data repositories—that advance representation, governance, and equity in genomic research.

Dr. Tsosie is a co-founder of the Native BioData Consortium, the first Indigenous-led biobank in the United States, and her work advances Indigenous data sovereignty and community-driven approaches to biomedical research.

Her scholarship examines how research design, data practices, and lived experience shape the quality, relevance, and impact of biomedical knowledge. She advocates for more equitable and community-embedded approaches to translational research, where participants are partners in shaping research questions, study design, and pathways to real-world health outcomes.

Generative deep learning for modeling single-cell differentiation dynamics: from prediction to interpretation

Abstract: Inferring the governing dynamics of differentiation that capture cell state evolution remains a central challenge in single-cell biology. Single-cell omics technologies resolve cellular heterogeneity at high resolution, but provide only static snapshots of continuous developmental processes. In this talk, I will describe two recent generative deep learning approaches to model continuous differentiation dynamics from static snapshot measurements.

First focusing on temporal predictions, I will describe CellPace, a model that learns developmental dynamics by leveraging a transformer-based temporal diffusion backbone. Across diverse mouse developmental lineages, CellPace achieves state-of-the-art performance in simulation, interpolation, and temporal forecasting of single-cell transcriptomic profiles. Beyond global statistics, CellPace learns and preserves fine-grained biological structure such as gene regulatory mechanisms, marker activation patterns, and spatial fidelity. Finally, I will show how this model can be generalized to multimodal data, jointly modeling transcriptomic and chromatin accessibility information.

Next focusing on model interpretability, I will describe Latent Space Dynamics (LSD). Inspired by thermodynamics and Waddington’s epigenetic landscape, LSD leverages a Neural ODE to jointly infer a compact cell state, a differentiable potential function governing developmental flow, and a local entropy term that quantifies cellular plasticity. Across diverse developmental systems, LSD accurately recovers lineage hierarchies, predicts fate commitment for unseen cell types, and outperforms existing trajectory inference approaches in directional accuracy. Moreover, in silico gene perturbations reveal how individual regulators reshape the landscape.

Together, these models showcase how recent advances in generative deep learning can be leveraged to learn single cell dynamics and provide models that enable temporal predictions and biological interpretations.

Dr. Amin Emad’s Biography

Dr. Emad is an Associate Professor of Electrical and Computer Engineering at McGill University and is affiliated with Mila (Quebec AI Institute), the Rosalind and Morris Goodman Cancer Institute, and the Victor Phillip Dahdaleh Institute of Genomic Medicine. Before joining McGill, he was a Postdoctoral Research Associate at the NIH KnowEnG Center of Excellence in Big Data Computing associated with the Department of Computer Science and the Institute for Genomic Biology at UIUC. He received his PhD in Electrical and Computer Engineering from University of Illinois at Urbana-Champaign (UIUC).

Dr. Emad’s research focuses on developing novel computational methods to study diverse biological systems. He develops computational approaches to model cellular dynamics and disease progression, to predict responses to perturbations, to infer gene regulatory networks, and to decipher protein-protein interactions, with applications in developmental biology and cancer precision medicine. In recent years, his work has centered on foundation models, geometric deep learning, generative AI, physics-informed neural networks, and causal AI for modeling complex biological systems.