Computational Stem Cell Research

A new blog from OpenAI and Retro Bio describes a custom AI model (“GPT-4b micro”) that can design better Yamanaka factors, the proteins used to reprogram mature cells into induced pluripotent stem cells. In 2006, Shinya Yamanaka discovered four proteins that could, together, “reprogram” skin fibroblast cells back into stem cells. He began with 24 transcription factors and removed them, one by one, until only four "essential" proteins remained: OCT4, SOX2, KLF4, and c-MYC. GPT-4b micro was trained on protein sequences, biological text, and tokenized 3D structures. (The actual training data has not been released.) This model “designed novel variants of the Yamanaka factors that achieve a 50x increase in reprogramming efficiency in vitro compared to standard proteins.” The blog post links to some prior efforts to engineer Yamanaka factors, but not many. Therefore, I built a table showing more studies in which scientists engineered efficient Yamanaka factors. One of these studies, not mentioned in the blog, reported an engineered Oct4 protein that raised “the efficiency of making mouse and human iPSCs more than 50-fold in comparison to” standard Yamanaka factors. It was published in 2011. Here are some of those prior efforts: 2011 — Researchers fused OCT4 to a transactivation domain. Adding this to Sox2, Klf4, and c-Myc boosted iPSC efficiency “more than 50-fold” in both mouse and human fibroblasts. 2016 — Mutant versions of KLF4 improved reprogramming efficiency and reduced unwanted differentiation. 2018 — Several groups showed that c-MYC, the most oncogenic of the four factors, could be replaced with safer alternatives without sacrificing efficiency. These are just a few examples. In reality, the literature is filled with rational engineering of Yamanaka factors, many reporting big jumps in efficiency. The difference between these prior efforts and the OpenAI/Retro work is not the goal, but the method. Previous groups used mutagenesis, fusions, or structure-guided design. These AI-designed proteins, by contrast, are far more distinct from natural Yamanaka proteins. The AI model is able to explore a “broader space” of possible designs, and that is really cool. It's something that I hope scientists will be able to build upon. Until the protein sequences are shared, though, it will be difficult to benchmark claims, because "efficiency" is defined so differently across experiments. Also, one of the reasons we even WANT better Yamanaka factors is to use them in therapeutics, such as for "partial reprogramming" efforts that aim to reverse aging inside of human organs. But in those cases, efficiency is not the biggest bottleneck: Delivery is. Still, the work is encouraging. It shows, as many other models have recently shown, that AI models can be repurposed for biology and be used to generate viable proteins that go beyond what evolution has otherwise sampled. A high-res table is available here: https://lnkd.in/eJ2pP9Wg

Current approaches used to track stem cell clones through differentiation require genetic engineering or rely on sparse somatic DNA variants, which limits their wide application. Here we discover that DNA methylation of a subset of CpG sites reflects cellular differentiation, whereas another subset undergoes stochastic epimutations and can serve as digital barcodes of clonal identity. We demonstrate that targeted single-cell profiling of DNA methylation at single-CpG resolution can accurately extract both layers of information. To that end, we develop EPI-Clone, a method for transgene-free lineage tracing at scale. Applied to mouse and human haematopoiesis, we capture hundreds of clonal differentiation trajectories across tens of individuals and 230,358 single cells. In mouse ageing, we demonstrate that myeloid bias and low output of old haematopoietic stem cells are restricted to a small number of expanded clones, whereas many functionally young-like clones persist in old age. In human ageing, clones with and without known driver mutations of clonal haematopoieis are part of a spectrum of age-related clonal expansions that display similar lineage biases. EPI-Clone enables accurate and transgene-free single-cell lineage tracing on hematopoietic cell state landscapes at scale. Interesting new paper detailing EPI-Clone, a computational method that analyzes DNA methylation at single-cell resolution, serving as a natural barcode to track stem cell clones without genetic engineering. By @Michael Scherer and larger team: https://lnkd.in/e3BMUhbB

“The first thing just worked,” Boris Power, head of OpenAI’s applied research team, told me ahead of the company's latest announcement. “That’s rarely the case in research. We were skeptical for a very long time.” Power was talking about GPT-4b micro, a protein-focused variant of its GPT-4o model that it built in collaboration with Retro Biosciences. The research shows how LLMs could be applied to life sciences research. In this case, by making variants on the famed Yamanaka factors that were more efficient in turning mature cells back into stem cells. Retro CEO "Joe Betts-LaCroix" expects to use some of these AI-made proteins in a preclinical research program, seeking to reprogram patient's cells. “Because reprogramming is in the loop, the timing and efficiency of it matter for the patient in terms of how many starting cells do you need, how long does the patient have to wait around,” Betts-LaCroix said. “It can work with canonical Yamanaka factors, but as we’re optimizing, we’re like, ‘Why?’ They’re worse.” My latest exclusive on the AI bio frontier at Endpoints News: https://lnkd.in/gU8_8ppf

In this newsletter, we explore the exciting integration of CRISPR lineage tracing with single-cell epigenomics, a powerful strategy to decode the origin and regulatory identity of individual cells. By combining CRISPR-based barcoding systems with chromatin accessibility or DNA methylation analysis at single-cell resolution, researchers can reconstruct lineage trees while simultaneously mapping epigenetic changes that drive cell fate decisions. This dual approach is changing the way we study complex biological systems such as development, regeneration, and cancer, revealing not only where cells came from but also how their regulatory landscapes evolved. It is also providing new insights into stem cell research and in vitro modeling, and offering a precise method to validate differentiation protocols. #CRISPRLineageTracing #SingleCellEpigenomics #LineageMapping #CellFateTracking #ATACseq #Methylome #ChromatinAccessibility #StemCellResearch #DevelopmentalBiology #RegenerativeMedicine #SyntheticBiology #SingleCellOmics #GeneEditing #Epigenetics #CSTEAMBiotech

AI-Powered Stem Cell Manufacturing: Unlocking the Potential of iPSCs for Regenerative Medicine The article discusses a new study by researchers at Northeastern University that proposes an AI framework for the mass manufacturing induced pluripotent stem cells (iPSCs) for regenerative medicine applications. iPSCs have the potential to transform into various specialized cell types and offer great promise for treating conditions like cancer, Alzheimer's, and Parkinson's and repairing spinal cord injuries. The researchers developed a modular "Biological System-of-Systems" (Bio-SoS) framework to model and predict the complex cellular metabolic and genetic processes involved in growing high-quality iPSCs at scale. The framework uses a combination of mechanistic models and interpretable AI to optimize iPSC cultivation parameters and ensure consistent cell product quality. The framework can model single-cell behavior and interactions within 3D iPSC aggregates grown in suspension bioreactors - a critical step for mass production. Incorporating interpretable AI allows the framework to learn and improve as more experimental data becomes available. This research is significant for regenerative medicine because it addresses a crucial bottleneck in realizing the full potential of iPSCs for regenerative medicine—the ability to reliably and efficiently manufacture large quantities of high-quality stem cells. The AI-powered Bio-SoS framework provides a foundational platform for understanding, predicting, and optimizing the complex biological processes involved in iPSC cultivation. This paves the way for the mass production of iPSCs required to support various clinical applications, drug development, and research. This technology could accelerate the translation of stem cell therapies from the lab to the clinic by enabling faster, more flexible, and robust iPSC manufacturing. This is vital for addressing unmet medical needs and expanding the reach of regenerative medicine to treat a wide range of debilitating diseases and injuries. Overall, this research represents a significant step forward in realizing the full potential of stem cell-based regenerative medicine. I am optimistic that we will hear more and more about AI and Regenerative Medicine.  JP https://lnkd.in/e6kp6WNV