AI summary by TwinMind
What If a Simple Blood Test Could Detect Cancer? | Hani Goodarzi | TED
TED
7m 3s
RNA Fundamentals and Cancer Detection Challenge
- Early cancer detection saves lives but challenge: finding small groups of rogue cancer cells among trillions of healthy cells
- RNA understanding increased due to COVID vaccines
- RNA transcribed from DNA, mRNA serves as protein synthesis template
- More mRNA typically = more protein production
Discovery of Orphan Non-Coding RNAs (OncoRNAs)
- New class of relatively small RNAs discovered
- Non-coding (don't code for proteins)
- Named "orphan noncoding RNAs" or "oncoRNAs" (OncoRs)
- Transform cancer detection approach from blood non-invasively
- Provide window into tumor biology itself
Cellular Genomic Machinery and Cancer Hijacking
- Every cell shares same genetic code but uses different "recipes"
- 200+ cell types with distinct roles (skin cells, neurons)
- Complex cellular machinery governs gene expression
- Each cell has 20,000 genes requiring precise expression regulation
- Cancer cells hijack this machinery:
- Increase expression of growth/spread genes
- Silence/downregulate cancer-suppressing genes
- Cancer cells "hack" original genomic recipe
Genomic Reprogramming in Cancer
- Consequence of genomic reprogramming = hallmark of cancer
- Genome parts normally silent in healthy cells become activated in cancer
- Direct consequence: birth of new RNA type only seen in cancer cells
- OncoRNAs absent in healthy cells
OncoRNA Mapping and Molecular Barcoding
- Years spent mapping cancer-emergent RNAs across human cancers
- OncoRNAs in given sample not random - tied to cancer type/subtype
- Collectively provide digital molecular barcode capturing cancer cell identity
- Barcode unique to specific cancer type/subtype
Blood Detection Mechanism
- OncoRNAs not confined to cancer cells
- Some packaged and released into blood (similar to healthy cells with other small RNAs)
- OncoRNAs only expressed in cancer cells + some enter bloodstream = blood detection opportunity
Partial Barcode Challenge and AI Solution
- Blood samples contain partial barcode only
- Only subset of OncoRNAs secreted from cancer cells to blood
- Even smaller subset reliably detectable in small blood volume
- Machine learning/AI reconstructs original tumor barcode from partial information
- Reconstructed barcode matched against OncoRNA catalog across cancers
- Enables: disease detection + type/subtype identification
- Growing catalogs allow deeper disease biology understanding
Clinical Application - UCSF Collaboration
- Preliminary study: 200 breast cancer patients
- OncoRNAs detect residual disease post-treatment
- Identifies patients needing additional treatment/monitoring after surgery
- Patients receive more treatment only when needed
Future Vision
- Next decade = decade of cancer screening
- Blood-based cancer detection major frontier
- AI + molecular barcodes of OncoRNAs enable:
- Precise and sensitive detection
- Accessible blood-based cancer detection
- Vision: not just hope but achievable reality
