This is a conversation I had with ChatGPT. My questions and comments should not be construed as reflecting any deeply held belief or perspective on my part. I have not reviewed any of the LLM responses for accuracy, so caution is advised before taking any action based upon the contents herein.
The genomic revolution has been a bit disappointing when it comes to applications to medicine. It turned out that there was only a small percentage of diseases that are caused by a single or small number of genes. There is now hope that AI will be able to unravel the complexity of diseases that may involve hundreds of genes. And that AI will be able to target these genes with therapies such as multi-location CRISPR editing to cure illnesses or reverse aging. However, if one is familiar with the cutting edge research of the Michael Levin laboratories on the role of bioelectricity in development, morphogenesis, regeneration, and cancer, there is a reason to be skeptical about the mainstream focus on gene editing or other therapies that focus on the molecular or genetic level. From Levin’s perspective, this is a very crude approach that mimics the way computers in the 1940s needed to be rewired at the hardware level to perform different operations. I find Levin’s approach to be much more elegant, in that it acts on the bioelectrical “software” layer of physiology that he and his team have demonstrated in hundreds of experiments. They have repeatedly shown that in the presence of the right electrical signals, the body knows how to make healthy tissues and organs without having to make any alterations to the genome.
ChatGPT said:
Thought for 15 seconds
Below is a high‑level survey of where each line of research stands in early 2025, why they sometimes talk past one another, and how they may ultimately converge.
1 Why “plain‑vanilla” genomics stalled for complex disease
- Polygenic architecture. Most common disorders (type‑2 diabetes, schizophrenia, atherosclerosis, many cancers) involve hundreds–thousands of loci of small effect, dispersed across regulatory DNA. Merely correcting one variant rarely moves the clinical needle.
- Missing layers. Epigenetics, 3‑D chromatin structure, microbiome, mechanical forces, and bioelectric states all modulate gene expression, so DNA sequence explains only part of phenotype.
- Regulatory headwinds. The first CRISPR therapy (for sickle‑cell disease) was approved in late 2024, but it targets a single well‑understood mutation. Extending that success to hundreds of sites per patient multiplies manufacturing, delivery, and safety hurdles. WIRED
2 What AI is already adding to gene editing
| Capability | 2023‑25 examples | Why it matters |
|---|---|---|
| Design entirely new Cas proteins | OpenCRISPR‑1 LLM‑generated editors broaden PAM compatibility and shrink cargo size. CRISPR Medicine | Smaller, multiplexable payloads. |
| Pick dozens of on‑target guides with minimal off‑target risk | Graph‑neural‑net tools now screen millions of gRNAs in silico in hours. ScienceDirect | Makes multi‑locus editing more realistic. |
| Plan combinatorial edits | Exon‑skipping predictor SkipGuide models >¹⁰³ guide/exon pairs to rescue dystrophin and other transcripts. Frontiers | Tackles polygenic or splice‑defect diseases. |
Bottom line: AI is turning genome surgery from artisanal to industrial, but the field still operates at the hardware layer Levin criticises.
3 Levin‑style bioelectricity: a “software” control layer
- Core idea: Cells treat resting‑membrane voltage, ion‑flux patterns, and gap‑junction connectivity as a distributed code that guides large‑scale anatomy, much like bits in a computer.
- Key 2024 data points
- Voltage‑gated modulation corrected dysplastic planarian tissue without DNA edits. Levin Lab
- Computational models showed tissues re‑establish normal electric patterns after perturbation—akin to error‑correction in software. Levin LabLevin Lab
- “Living robots” (Xenobots 3.0) self‑repair and perform tasks based solely on bioelectric cues. Tufts Now
- Implication: If you can overwrite the pattern (e.g., with ion‑channel drugs, optogenetics, or external fields), the genome’s existing circuits execute the build/repair instructions for you.
Todd Luger's Blog 