Digital Twin to Model the Effect of Spaceflight on the Retina

AWG Projects AIML AWG Open Projects
, ,
122

Hello! Sorry for not attending the past meetings. I had serious commitments. Hopefully, I would be able to attend the upcoming meetings.

123

2026/5/1 - Meeting Recap:

chat_record_2025-05-01.pdf (87.1 KB)
Quick recap

The meeting focused on discussing current research projects and data access challenges in space biology. Participants shared updates on their work, including Alireza’s interactive network dashboard for human system risks, Suja’s project on merging health and space technology data, and Nozipho’s proposal for using machine learning to predict genomic instability from radiation exposure. The group discussed data access limitations, particularly regarding human astronaut data, with Lauren and Fathi providing insights about international data partnerships and the TRISH program’s restrictions. The conversation included detailed feedback on Nozipho’s research approach, with Fathi suggesting focus on DNA damage response pathways, p53 signaling, and oxidative stress markers. The meeting also covered Suja’s work on creating a sovereign infrastructure for health data and her plans to present at the MedTech Conference in Florida.

Next steps

  • Max: Coordinate with Jian to schedule a future meeting to present updates on human research work in Japan.

  • Suja: Present a detailed introduction and slides about her project after returning from the MedTech Conference in Florida.

  • Jian: Share the transcript (or recording) of the meeting on the forum for group access.

  • Nozipho: Review feedback and references shared during the meeting, refine her project proposal, and update the group in a couple of weeks on her progress and next steps.

  • Jian: Share the proposal written for TRISH Expand access with the group for feedback.

  • Alireza: Introduce Robert Reynolds to the group if there is a need for collaboration on clinical/case scenarios.

  • All interested members: Consider submitting a proposal for the TRISH Expand data access call (encouraged by Alireza and Jian).

Summary

Iran Conditions and Research Updates

Jian and Alireza discussed current conditions in Iran, where Alireza reported high inflation but noted that people have adapted to it. Alireza expressed concern about the potential for total war with other countries as the main source of anxiety for Iranians. Fathi informed Alireza that he had provided a recommendation for Alireza’s application to the Blue Marble Science Institute, a non-profit research organization that offers networking opportunities and proposal writing possibilities.

Project Updates and Introductions Meeting

The meeting served as an informal check-in where participants introduced themselves and discussed ongoing projects. Nozipho was invited to present her project idea at a later time, though the specific details were not discussed. Atul mentioned working on a manuscript about fungus or bacteria research, and Max, who was joining from Japan at 2 AM, agreed to schedule a future discussion about his human research work. Suja provided a detailed introduction of her background in electrical engineering and software development, explaining her current project focused on building an infrastructure for digital twins using medical records, which she plans to present at the MedTech Conference in Florida.

Rural Medicine and Space Integration

Suja discussed the potential for integrating rural medicine, health, and space technology, highlighting a $3 trillion industry opportunity over the next 10 years. She emphasized challenges with data quality and privacy, particularly in open science and citizen science initiatives, and expressed interest in developing solutions to address these issues. Jian acknowledged the difficulties in accessing astronaut data and mentioned ongoing efforts to obtain deeper access through requests to relevant agencies.

OSDR Training and Data Collaboration

The group discussed the helpfulness of OSDR training modules, with Suja sharing her positive experience taking three classes that covered data navigation, formatting, policies, and research project structuring. Jian noted that data access and analysis skills are fundamental to their projects and mentioned plans to organize more focused coding sessions and purpose-built pipelines to help lower barriers for team members. Alireza raised questions about accessing data from other countries and inquired about potential collaboration goals with countries like Japan, China, and Australia to address data constraints and privacy concerns.

International Space Biology Data Repositories

Lauren and Fathi discussed international data repositories for space biology research. They explained that OSDR is the primary international database, with data from JAXA and an upcoming agreement with the European Space Agency to share human analog spaceflight data through Gen Lab. Fathi mentioned that JAXA’s astronaut data release was facilitated by Masafumi Muratani, who used an N6 approach to address identification issues. They also noted that while OSDR is open science, TRISH offers access to private astronaut data through a proposal-based system rather than open access.

Data Analysis Proposal Discussion

Jian submitted a proposal to explore data analysis opportunities but found the requirements restrictive, requiring specific hypotheses rather than open-ended exploration. Fathi agreed that the approach was too constrained, particularly requiring proposals to address HRP risk factors rather than broader topics like regenerative medicine and aging. Jian shared his specific hypothesis about using wearable sensor data to predict short-term physiological effects of spaceflight stress, though he acknowledged it was not scientifically sound but rather an exploratory approach to test what might be possible with the data.

Wearable Data Collection for Space

The team discussed wearable data collection from suborbital flights and potential terrestrial alternatives. Jian mentioned having access to kinesiology lab sensors and collaborators for processing data sets. Alireza shared information about a sleep risk section in a DAG catalog, highlighting the importance of sleep quality in space missions. Jian explained Alireza’s interactive dashboard based on NASA’s DAG paper, which allows users to explore causal relationships between various space mission factors.

NASA Graph Platform Integration Discussion

Jian presented an interactive platform based on a NASA document related to human systems risk and cyclic graphing. Alireza shared the link to the resource and mentioned potential applications in processing time series data for sleep measures or astronaut workload. Jian and Alireza discussed exploring ideas for integrating this graph-based approach into their CDSS project, with Alireza offering to collaborate with Robert Reynolds on clinical scenarios. The conversation ended with Jian preparing to transition to the next topic and inviting Nosy to share their ideas.

Space Radiation DNA Repair Research

Nozipho presented a research concept using machine learning and AI models to predict genomic instability caused by radiation in spaceflight conditions, focusing on DNA damage and its long-term effects. Jian suggested studying DNA repair mechanisms instead of damage directly, referencing a recent talk on longevity in mammals and suggesting the use of gene expression data from OSDR to identify DNA repair genes activated in space mice. The group discussed different perspectives on the relationship between stress, DNA repair, and aging, with Fathi highlighting inflammation as a key factor in aging. No specific decisions or action items were outlined, but the discussion provided insights into potential research directions and data sources for further exploration.

Radiation Effects and DNA Damage

Fathi discussed the biological effects of radiation, explaining that it can be divided into direct effects (like high-energy particles causing DNA damage) and indirect effects (through reactive oxygen species). He suggested focusing on DNA damage response pathways, including p53 signaling and ATM pathways, as well as measuring oxidative stress as a proxy for indirect damage. The team agreed to explore both transcriptomic data and potentially complementary functional assays like comet assays and FISH for validating their AI and machine learning models. Jian mentioned he would share the meeting transcript on the forum for further discussion.

3 Likes
124

Hi Jian, this is really interesting—I’d love to contribute.

1 Like
125

Hi @jgong, this is a great initiative. I’m a practicing physician with significant medical research and AI experience. I’m currently working on digital twins for clinical support. Would love to contribute to your digital twin model for retina if you need any clinical perspectives.

3 Likes
126

Hi, I find this project really interesting. Since I work on eye myself, I would love to contribute in this project

1 Like
127

Part of this just sub-group on the @AIMLawg – just released this pre-print!

https://doi.org/10.64898/2026.04.22.720210

3 Likes
128

This might be a long post (sorry), but we successfully build a OSDR chatbot that can ingest OSDR metadata and answer open-ended questions:

The first few responses were helpful and I will share them here. We were investigating all open datasets available and related to eye/SANS.

OSDR Eye/SANS ChatBot
Model: gemma4:latest | Studies loaded: 15
Type ‘quit’ or Ctrl+C to exit, ‘studies’ to list loaded studies

You: What data do you have about SANS on OSDR?

Assistant: Based on the provided metadata, I have access to a comprehensive set of data regarding the neuro-ocular effects observed in spaceflight, which directly relates to the understanding and investigation of Space Associated Neuro-ocular Syndrome (SANS).

The data distinguishes clearly between direct human astronaut studies (though no raw human clinical data is listed, the context is based on those findings), ground-based mechanical models using rats, and actual spaceflight experiments conducted on mice.

:rocket: I. Data from Actual Spaceflight Studies (Mice – Mus musculus)

These studies are the most direct evidence of SANS mechanisms, utilizing International Space Station (ISS) animals (Mus musculus). The general trend observed is that microgravity induces molecular and structural changes in the retina and optic nerve head.

OSDR ID Target/Organism Assay Type(s) Factor/Mission Key Finding Area Relevant to SANS
OSD-397 Retina (Mice) RNA-seq, DNA Methylation (RRBS) Spaceflight (SpaceX-4) Investigates global gene expression and epigenetic changes in the retina following space exposure.
OSD-255 Retina (Mice) RNA-seq Spaceflight (SpaceX-12) Determines how the environment induces oxidative damage and subsequent shifts in gene expression profiles within the retina.
OSD-758 Retina (Mice) RNA-seq Spaceflight (SpaceX-27) Compares transcriptomic responses to spaceflight versus altered gravity, detailing molecular adaptation/disruption mechanisms.
OSD-557 Eye / Photoreceptor (Mice) Histology, Micro-CT, Molecular Imaging Spaceflight (SpaceX-12) Provides structural data on photoreceptor integrity and oxidative stress markers (e.g., 4-HNE assay).
OSD-583 Both Eyes (Mice) Intraocular Pressure (IOP) / Tonometry Spaceflight (SpaceX-12) Functionally measures IOP changes, which is a critical biomarker for glaucoma risk and neuro-ocular pressure imbalance.

:anchor: II. Data from Ground-Based Analog Models (Rats – Rattus norvegicus)

These studies use mechanical analogs (like hindlimb unloading or head-down tilt) to simulate the physical stresses of microgravity on the vascular, structural, and neurological systems.

OSDR ID Target/Organism Assay Type(s) Model/Factor Key Finding Area Relevant to SANS
OSD-679 Eye (Rats) IOP, OCT, MRI, Tonometry Hindlimb Unloading Comprehensive assessment package including functional metrics (IOP/Tonometry), structural imaging (MRI), and retinal detail (OCT). Best for overall physiological testing.
OSD-680 Optic Nerve (Rats) Tomography / Bruker Biospec 7T Hindlimb Unloading Focuses specifically on the structure of the optic nerve, assessing potential axonal or tissue changes due to mechanical unloading.
OSD-681 Subdural Space/ICP (Rats) ICP, Body Temp Monitoring Ground Model Tracks systemic pressures (Intracranial Pressure) which are highly correlated with retinal and optic nerve compression associated with SANS.

:microscope: III. Summary of Available Assay Types and Biomarkers

My access to the metadata covers a wide range of techniques necessary for characterizing SANS pathophysiology:

A. Functional Measurements

  • Intraocular Pressure (IOP) & Tonometry: Directly measured using TonoLab-ICARE or generalized tonometric measurements, crucial for assessing globe geometry and potential glaucoma risk ([OSD-583], [OSD-679]).
  • Intracranial Pressure (ICP): Measured via telemetry in ground models, helping to understand the systemic pressure component of neuro-ocular changes ([OSD-681]).

B. Structural Imaging & Anatomy

  • Optical Coherence Tomography (OCT): Used to visualize detailed retinal layer integrity and changes at the optic nerve head in animal models ([OSD-679]).
  • Magnetic Resonance Imaging (MRI) / Tomography: Provides macro-level views of the optic nerve, subtidal space, and overall ocular structure ([OSD-680], [OSD-679]).
  • Micro-CT: Used to assess bone microstructure changes that may accompany long-term spaceflight analogs ([OSD-557]).

C. Molecular & Cellular Insights

  • Transcriptomics (RNA-seq): High-throughput screening of gene expression profiles in the retina, allowing researchers to identify pathways altered by microgravity or oxidative stress ([OSD-397], [OSD-255], [OSD-758]).
  • Epigenetics: Reduced Representation Bisulfite Sequencing (RRBS) data provides insight into DNA methylation patterns, suggesting altered gene regulation independent of DNA sequence change ([OSD-397]).
  • Histology & Immunohistochemistry: Allows visualization and quantification of specific cellular components (e.g., photoreceptors) and oxidative stress indicators like 4-HNE or Prussian Blue staining for iron metabolism ([OSD-557]).

Disclaimer: I do not have access to the raw data files themselves, only the metadata summarizing the experimental parameters, organisms, tissues, assays, and missions associated with the studies.

You: I am trying to design a new study. Where are the largest gaps of understanding about SANS that these data can potentially answer?

Assistant: Based on a detailed analysis of the available metadata, the existing studies provide robust foundational knowledge regarding the mechanisms of SANS in microgravity (stressors, structural changes, molecular shifts).

However, the biggest gaps lie not merely in collecting more data, but in integrating systemic physiological measurements and translating fundamental biological findings into actionable clinical interventions.

Here are four major gaps in understanding SANS that these existing datasets could potentially help answer through innovative study design:

:microscope: Gap 1: The Interplay Between Cerebrospinal Fluid (CSF) Dynamics and Retinal Health

The current studies measure Intracranial Pressure (ICP) ([OSD-681]) and general IOP, but they often treat these as separate inputs. SANS is fundamentally a problem of fluid dynamics within the enclosed space of the skull, impacting both the optic nerve sheath and the retina.

  • The Gap: We lack high-resolution data on how minute changes in cerebrospinal fluid (CSF) flow velocity or composition directly relate to retinal layer integrity in vivo. Is it simply pressure, or is it mechanical shear stress or subtle chemical exchange?
  • Potential Study Design: A combined model study utilizing the ICP/Tonometry infrastructure ([OSD-679], [OSD-681]) but adding assays for CSF biomarkers (e.g., neurofilament light chain via proteomics) and measuring deep retinal vascular architecture using advanced OCT techniques, correlated with changes in orbital volume dynamics.
  • Goal: To pinpoint the critical threshold of fluid dynamic imbalance that initiates damage before measurable structural decline occurs.

:rocket: Gap 2: Translating Molecular Changes to Functional Decline (The Phenotype Link)

We have excellent molecular data (RNA-seq, epigenetics from [OSD-397], [OSD-255]), but the link between these profound genomic alterations and observable functional vision loss is poorly defined.

  • The Gap: We know that certain genes are down-regulated or methylation patterns change ([OSD-100]). But which specific molecular cascade, when disrupted, leads to measurable impairments in visual acuity, retinal ganglion cell (RGC) survival, or macular function?
  • Potential Study Design: A longitudinal, multi-omics approach combining transcriptomics with single-cell resolution assays. Instead of bulk RNA-seq on the retina, focus on single-nucleus sequencing specifically targeting RGCs and Retinal Pigment Epithelium (RPE). This would allow us to differentiate cell-specific susceptibility and identify key signaling pathways affected by microgravity stressors.
  • Goal: To create a functional molecular “signature” of early SANS that can be detected long before structural or physiological decline is evident on imaging.

:test_tube: Gap 3: Efficacy and Timing of Countermeasures (The Intervention Test)

All current studies are largely observational—they report the changes after exposure to microgravity or analog stressors. We have limited data testing if specific interventions can prevent the damage in real time.

  • The Gap: We do not know which countermeasures (e.g., fluid loading, dietary supplements, targeted gene therapy analogs) would most effectively disrupt the pathological cycle of SANS. Is it restoring IOP, maintaining CSF volume, or mitigating oxidative stress?
  • Potential Study Design: A controlled intervention study based on a robust ground-based analog model ([OSD-679]). The design must test multiple countermeasures (e.g., countermeasure X vs. Y) against the established microgravity challenge. Assays would then focus heavily on functional endpoints like maintaining normal RPE viability or preventing optic nerve axonal degeneration—data not explicitly highlighted in the existing metadata but crucial for translational research.
  • Goal: To move beyond correlation and establish causality: demonstrating that Countermeasure A prevents Marker B, leading to outcome C (e.g., Preventing reduced photoreceptor integrity $\rightarrow$ Maintaining normal OCT structure).

:counterclockwise_arrows_button: Gap 4: The Transition From Ground Analog to Spaceflight Reality (Model Fidelity)

While [OSD-679] and [OSD-681] are excellent ground analogs, the physiological complexity of spaceflight involves unique elements that may not be perfectly simulated by simple HUL or tilt protocols.

  • The Gap: How do confounding factors like sustained radiation exposure (which affects cell DNA) or chronic psychological stress interact with microgravity on ocular tissues? We need models that combine mechanical unloading with systemic stressors.
  • Potential Study Design: An advanced model integrating multiple simultaneous stresses. For example, utilizing a hindlimb unloading animal system but adding simulated radiative exposure protocols alongside the standard imaging and molecular assays (OSD-679). This would simulate the full burden of deep space travel, rather than isolating single variables.
  • Goal: To create an integrated understanding of SANS etiology that moves beyond “unloading” to encompass the comprehensive bioburden of extended missions in orbit.
2 Likes
129

This looks great Jian! Is the chat bot using RAG in the back end?

130

Hi,

Thank you for sharing the meeting notes. Unfortunately, I couldn’t attend due to the early-morning scheduling conflict.

The emphasis on explainability, reasoning visualization, and user-centered design aligns closely with the governance and human-factors considerations we’ve been discussing earlier on.

I also noticed the references to Google/DeepMind medical reasoning architectures. I would be happy to help review and map those concepts into our CDSS framework, particularly around:

  • AI reasoning and orchestration

  • RAG and knowledge management

  • Explainability and traceability

  • Evaluation and benchmarking

  • AI safety and governance

  • Human-in-the-loop decision support

For those who may not have seen it previously, I had also created an initial repository structure proposal to help organize the architecture discussions and implementation workstreams:
NASA-CDSS-for-Long-Duration-Spaceflight: GitHub - aka79/NASA-CDSS-for-Long-Duration-Spaceflight: Clinical Decision Support System (CDSS) for Long-Duration Spaceflight · GitHub

Looking forward to getting aligned with everyone —I’m happy to contribute wherever I can be most useful (Note: I’ll be away in July for some travels).

1 Like
131

Hello @jgong ! Just became a part of this group and I found this project to be really interesting. I would love to contribute in any way possible. I saw that there is a coding session scheduled tomorrow. Could you let me know how to get involved?

132

Hi Dr. Gong @jgong, I recently joined this group and found this project very interesting. I’d love to contribute and get involved. I noticed there’s a coding session tomorrow, could you let me know how I can participate?

133

Hi Dr. Gong, @jgong

I recently joined the AI/ML AWG and found the Digital Twin project really exciting, especially given my interests in computational biology, neuroscience, and AI.

I also reached out to you through chat to introduce myself and learn more about ways I might get involved. Whether it’s helping with data processing, literature review, coding, or other aspects of the project, I’d be happy to support the team and learn from everyone involved.

Looking forward to contributing!

Sincerely,

Zarif Ula

zarif.ula1@gmail.com

2 Likes
134

@kavyabhand @aadijoshi @ZarifUla you are all welcome to join our meeting today at 10am PST. Thank you for the interest!

1 Like
135

6/26/2026 - Digital Twin subgroup meeting notes

Hi all, we did a detailed deep dive/demo into the Biomi interface in today’s meeting. Overall a good meeting and a lot to reflect on. Please find useful links below. A bit side-tracked but I think it is important. Next week we will get back to projects and tasks.

Jian.

Meeting summary

Quick recap

This meeting focused on introducing new members to the space biology project and demonstrating a new AI platform called Biomi for biological data analysis. Jian led the session by welcoming new participants including Kaushik from Pune, India, Kavya from Maharashtra, Aadi from Pune, and others joining from various locations around the world. The group then explored the Biomi platform’s capabilities, with Jian demonstrating how it can search and analyze data from the OSDR (Open Space Data Repository), specifically examining datasets related to the Inspiration4 mission. The platform showed impressive ability to pull API documentation, metadata, and provide detailed analysis of gene expression and radiation effects on specific genes. Participants discussed the advantages of specialized bioinformatics tools versus general-purpose AI like ChatGPT, with concerns about validation and publication requirements. The conversation ended with an agreement for team members to explore the platform independently and continue developing workflows for using AI tools in their space biology research.

Next steps

Jian

Collaboration

  • All members: Explore the Biomi.bio platform (or similar AI/omics analysis tools) using their own research questions/workflows and prepare to discuss findings and optimal approaches in future meetings.
  • New members (Kaushik, Kavya, Aadi): Send Jian their email addresses to be added to the calendar list for meeting schedule updates.
  • All interested members: Test the ability of Biomi.bio (or similar tools) to process raw omics data (e.g., RNA-seq, protein folding) and report back on functionality, limitations, and validation for research/publication purposes.
  • All members: Reflect on how to effectively frame questions, use prompts, follow up, and organize information when working with AI tools, and be prepared to discuss best practices in upcoming meetings.

Summary

New Team Member Introductions Meeting

Jian welcomed new participants to the meeting and facilitated introductions. Alireza shared updates about his recent travel to Iran and discussed potential diplomatic developments. Kaushik introduced himself as an AI engineer from Pune, India with experience in healthcare startups and multiomics, who is preparing to submit research on AI governance to the AAAI conference. Kavya joined the group from Pune, Maharashtra, and shared her background in computer science with focus on machine learning, deep learning, and cybersecurity, expressing interest in contributing to the bioinformatics and neuroscience aspects of the project.

Global Team Introductions Meeting

The meeting included introductions from team members across different locations, including Aadi from Pune, India, who is a junior CS student with neuroscience interests and open source experience; Ayse from Germany discussing heat wave challenges affecting lab operations; Atul from LA; and Fathi from Korea joining at 2:14 AM. The conversation focused primarily on personal introductions and location updates, with Jian noting the diverse global participation in the project.

Biomarker Simulation and AI Platform

Alireza discussed a new paper from Afshin Bashti that uses a simulation approach to find biomarkers related to radiation in the OSDR dataset before validating results in real spaceflight data. Jian introduced a new AI interface called biomi.bio that can help with various biological data analyses including genomics, proteomics, and RNA sequencing, and mentioned that Ryan Scott has been using it with OSDR API to train foundation models. The team was invited to explore the platform together, with Jian suggesting they could use it to analyze gene expression data and run specific analyses on their current projects.

ChatGPT for OSDR Data Analysis

The team discussed using OSDR data to test whether ChatGPT could automatically process raw data like RNA-seq. Ritika explained that while general-purpose AI like ChatGPT produces less accurate results for biological data analysis compared to specialized applications, they would still test the capability. Jian demonstrated the OSDR interface which contains curated tools and databases including AlphaFold, Protein Data Bank, and gene ontology resources that could be useful for their analysis.

AI Tool for OSDR Analysis

The team demonstrated and discussed a new AI tool’s capabilities for analyzing OSDR (Open Science Data Repository) datasets, particularly focusing on the Inspiration4 mission data. Jian showed how the tool successfully identified and analyzed 10 relevant datasets, extracted metadata, and provided detailed information about gene effects from radiation exposure. The group discussed the tool’s underlying large language models, with Kaushik confirming it uses different LLMs for different agents including Gemini Flash 2.5 for light tasks. The team agreed to explore the platform individually and continue discussions about integrating AI tools into their research workflow for digital twin development and space-related studies.

1 Like