Biology subjects

Koo, S.

Publications and source records attributed to Koo, S..

3 recordsLinked to original sources

bioRxiv2026

A human developmental and adult brain atlas benchmarks dopaminergic stem cell models and cell therapy candidates

Parkinsons disease (PD) is characterized by the progressive loss of midbrain dopaminergic (mDA) neurons1. Stem cell-derived mDA neurons hold promise for disease modelling2,3 and are currently in clinical trials for cell replacement therapy4-6. However, systematic benchmarking has been limited by the lack of a unified high-resolution reference and methods that quantify incomplete or mixed lineage specification in vitro7. We establish a single-cell and spatial atlas of the human developing diencephalon-midbrain-hindbrain axis resolving 93 cell subtypes, including 39 lacking prior single-cell characterization and 4 entirely novel populations. Using this atlas as a reference, we integrate 19 hPSC-derived mDA datasets, both published2,8-25 and unpublished, to build the Human Dopaminergic Neural Atlas (HDNA) spanning 2D, 3D, and graft models, including those used in clinical trials. To classify cells and quantify lineage fidelity, we develop CapybaraBrain, a marker-driven non-negative decomposition framework that assigns each cell continuous identity scores across all 93 developmental programs, enabling systematic discrimination of discrete, transitioning, and cross-lineage hybrid states26. We uncover a pervasive landscape of off-target populations reflecting relaxed transcriptional boundaries in vitro, including a previously unrecognized TH-PITX2 midbrain neuronal population, and we validate atlas-predicted latent lineage plasticity through inducible genetic fate mapping in mouse models. We further define maturation-associated transcriptional programs by harmonizing adult mDA subtype atlases, revealing that dopaminergic identity and maturation are partially decoupled across protocols. Finally, projecting PD patient-derived tri-cultures onto the HDNA uncovers genotype- and cell-type-specific transcriptional dysregulation. Together, these integrated atlases and computational framework establish a unified standard for benchmarking differentiation fidelity, exposing off-target states, and guiding next-generation PD models and cell therapies.

Developmental Biology↗

bioRxiv2024

Bone morphogenetic protein (BMP) signaling determines neuroblastoma cell fate and sensitivity to retinoic acid.

Retinoic acid (RA) is a standard-of-care neuroblastoma drug thought to be effective by inducing differentiation. Curiously, RA has little effect on primary human tumors during upfront treatment but can eliminate neuroblastoma cells from the bone marrow during post-chemo consolidation therapy--a discrepancy that has never been explained. To investigate this, we treated a large cohort of neuroblastoma cell lines with RA and observed that the most RA-sensitive cells predominantly undergo apoptosis or senescence, rather than differentiation. We conducted genome-wide CRISPR knockout screens under RA treatment, which identified BMP signaling as controlling the apoptosis/senescence vs differentiation cell fate decision and determining RAs overall potency. We then discovered that BMP signaling activity is markedly higher in neuroblastoma patient samples at bone marrow metastatic sites, providing a plausible explanation for RAs ability to clear neuroblastoma cells specifically from the bone marrow, seemingly mimicking interactions between BMP and RA during normal development.

cancer biology↗

bioRxiv2020

Deep Metric Learning Network using Proxies for Chromosome Classification and Retrieval in Karyotyping Test

In karyotyping, the classification of chromosomes is a tedious, complicated, and time-consuming process. It requires extremely careful analysis of chromosomes by well-trained cytogeneticists. To assist cytogeneticists in karyotyping, we introduce Proxy-ResNeXt-CBAM which is a metric learning based network using proxies with a convolutional block attention module (CBAM) designed for chromosome classification. RexNeXt-50 is used as a backbone network. To apply metric learning, the fully connected linear layer of the backbone network (ResNeXt-50) is removed and is replaced with CBAM. The similarity between embeddings, which are the outputs of the metric learning network, and proxies are measured for network training. Proxy-ResNeXt-CBAM is validated on a public chromosome image dataset, and it achieves an accuracy of 95.86%, a precision of 95.87%, a recall of 95.9%, and an F-1 score of 95.79%. Proxy-ResNeXt-CBAM which is the metric learning network using proxies outperforms the baseline networks. In addition, the results of our embedding analysis demonstrate the effectiveness of using proxies in metric learning for optimizing deep convolutional neural networks. As the embedding analysis results show, Proxy-ResNeXt-CBAM obtains a 94.78% Recall@1 in image retrieval, and the embeddings of each chromosome are well clustered according to their similarity.

bioengineering↗