Blood-Based Early Detection of Non-Small Cell Lung Cancer Using Orphan Non-Coding RNAs
Mehran Karimzadeh, Jeffrey Wang, Aiden Sababi, Dare Afolabi, Ti Lam, Alice Huang, Diana Corti, Kristle Garcia, Seda Kilinc, Allen Zhao, Jeff J Wang, Taylor Cavazos, Patrick Arensdorf, Kimberly Chau, Helen Li, Hani Goodarzi, Lisa Fish, Fereydoun Hormozdiari, Babak Alipanahi
All authors are employees, consultants, or shareholders of Exai Bio Inc.
Debunking the myths about RNA
Cell-free RNA is stable and resilient
cfRNA in biofluids can be protected by extracellular vesicles and proteins
Tumor-derived cfRNA is abundant in the blood
As opposed to DNA, tumor cells actively secrete RNA into the blood
cfRNA is more informative than cfDNA
Tumor-derived RNA can reveal biologically relevant signatures of the underlying cancer
Orphan non-coding RNAs are novel biomarkers for cancer detection
Fish, L. et al, Nature Medicine, Vol 24, November 2018, pg. 1743-1751
oncRNAs + AI enable cancer detection
Each oncRNA is akin to a single pixel in an image. And similar to a single cfDNA methylation or fragmentation locus, a single oncRNA is not very informative
When the signal in thousands of oncRNAs is aggregated and distilled using AI, it informs on the presence and the biology of the tumor, even at an early stage
Early detection of NSCLC using oncRNAs in blood
NSCLC study cohort design
Training (n=369) and Test (n=171) cohorts are independently designed, procured, processed and analyzed
NSCLC and non-cancer subjects were comparable with regards to age, sex and BMI
Study enrollment emphasized earlier detection of NSCLC
Model development and evaluation
Model is based on a catalog of 80,000 distinct oncRNAs discovered from TCGA
RNA is isolated from <1 mL serum and the prepared libraries are sequenced at a depth fewer than 20 million 50-bp single-end reads
Model is an ensemble of logistic regression models
Training performance metrics are computed using 5-fold cross-validation
After locking the model, we applied it to an independent test cohort
oncRNAs + AI accurately detects NSCLC
oncRNAs + AI have high sensitivity to detect early stage NSCLC
oncRNAs + AI have high sensitivity to detect the smallest tumors
Summary
Unique RNA & AI technology
Stable, abundant and specific to cancer
Large catalogue enables detection of cancer patterns
Tissue-derived oncRNA signature is reliably detected in blood
Powerful results in NSCLC
Accurately detects NSCLC
High sensitivity by clinical stage
High sensitivity to detect smallest tumors
One universal assay with broad clinical applications across multiple cancers
Highly efficient technology platform
Single assay supports rapid product development
Acknowledgements
This presentation is made possible thanks to the following people:
