Retro Biosciences publicly describes work on haematopoietic stem-cell reprogramming, tissue reprogramming and chemical rejuvenation. These are organisation-level research efforts with different delivery and development problems. The public portfolio also includes autophagy and microglia work, which should be tracked separately.
Retro collaborated with OpenAI on protein design. In August 2025, the collaborators reported that a protein-focused experimental model generated variants of SOX2 and KLF4, two transcription factors used in reprogramming. Retro tested the sequences in laboratory cells.
The result opens an engineering question: can changing a reprogramming protein improve useful control over cell state? Marker expression gives an early answer about one experimental process. A therapeutic application would require a defined cell product or delivery system, consistent manufacture and evidence about its behaviour over time.
- Propose factor sequencesComputational design supplies candidates.
- Test the factors in cellsMeasure the actual response and cell identity.
- Define the product objectiveTransplantable cells and tissue gene therapy need different evidence.
A pluripotency-marker result does not establish restoration of an existing tissue. Sources for this account.
The mechanism
Transcription factors interact with DNA and other proteins to regulate gene expression. Altering their amino-acid sequence can change those interactions. A model can propose sequences, but an experiment must establish what the proteins do in cells.
The collaboration reported improved expression of pluripotency markers with engineered factors. Full pluripotency creates cells capable of producing many specialised cell types. A programme aiming to rejuvenate an existing tissue must also preserve the useful organisation and role of its cells. These are distinct objectives with different success criteria.
Retro's public science description includes producing rejuvenated blood-forming stem cells for transplantation and developing gene therapies for tissues. The first approach would involve preparing a cell product outside the body; the second would require controlling expression in the intended cells within a tissue. The current public pages do not provide a complete candidate-level evidence package for either approach.
Human evidence
The reported experiments used human cells, including fibroblasts and mesenchymal stromal cells. They provide laboratory evidence. No human clinical reprogramming study or outcome was verified for the specific programme covered here.
Retro's homepage states an ambition to add ten healthy years to human life. That is a company goal. The inspected sources do not establish a measured lifespan gain in people. Any clinical progress in another Retro programme, such as autophagy, would not automatically change the stage of its reprogramming work.
Studies and experimental evidence
GPT-4b micro reprogramming collaboration
- Design
- Company-reported protein-design and cell experiments
- Population or model
- Human fibroblasts and mesenchymal stromal cells
- Sample size
- Donor and experiment counts differ by assay; three donors specified for one stromal-cell experiment
- Comparison
- Natural reprogramming factors and assay controls
- Duration
- Cell assays over days to weeks
- Outcome
- Pluripotency markers, derived-cell properties and damage-associated markers
Collaborators reported improved marker expression using engineered SOX2/KLF4
Collaborator disclosure; human cells are not treated participants; independent replication needed
Funding and interests: OpenAI/Retro collaboration; disclosure notes Sam Altman's Retro investment
Partial reprogramming scientific foundation
- Design
- Cell and mouse experiments
- Population or model
- Mouse and human cells; premature-ageing and injury mouse models
- Sample size
- Varies by experiment
- Comparison
- Experiment-specific controls
- Duration
- Short and cyclic reprogramming schedules
- Outcome
- Age-associated cell measures, regeneration and premature-ageing mouse survival
Brief factor expression improved selected measures in these models
Contextual academic work; not evidence that a Retro candidate has reproduced the result
Funding and interests: Academic research; see primary paper disclosures
Development history
Engineered factors disclosed
OpenAI and Retro report their model-to-laboratory protein-engineering work.
Unresolved questions
- Will engineered factors improve the desired specialised-cell function as well as marker expression?
- Which Retro reprogramming effort will yield a named therapeutic candidate?
- Can the cell-state change be controlled reproducibly without unwanted differentiation?
- Will independent laboratories reproduce the reported sequence effects?
- How will delivery and manufacture affect the performance of the engineered proteins?
Retro has not disclosed a dated clinical milestone for this reprogramming programme in the sources listed here. A full experimental report with accessible methods and sequences would allow closer assessment of the cell findings. A named candidate would then need appropriate animal and clinical testing.
Sources
- Company disclosure · 22 Aug 2025Accelerating life sciences research with Retro Biosciences
Collaborators' experimental report; AI role documented, clinical efficacy not established.
Checked 4 Oct 2026 - Company disclosure · Date not statedRetro Biosciences science portfolio
Organisation-level objectives; no complete named reprogramming candidate dataset.
Checked 4 Oct 2026 - Company disclosure · Date not statedRetro Biosciences mission and strategy
Ten-year healthspan figure is an ambition, not an achieved result.
Checked 4 Oct 2026 - Paper · 15 Dec 2016In Vivo Amelioration of Age-Associated Hallmarks by Partial Reprogramming
Academic field foundation; predates Retro and is not a Retro candidate result.
Checked 4 Oct 2026
Dr T Smith, organic chemist and science educator. Review date: 4 Oct 2026. Report a correction.