Although receiving a selfie from beyond the grave might sound disturbing, posthumous snapshots from dying cells in the form of the cell-free DNA (cfDNA) they release can help prevent us from entering an early grave. Giving Snapchat a run for its money is a new program called CelFiE (CELl Free DNA Estimation via expectation-maximization), which characterizes cfDNA via a simple blood draw to reveal potential biomarkers for disease.
While a typical cfDNA snapshot is primarily composed of DNA from blood and liver cells, this image is distorted when distinct cell populations die differentially during pregnancy, degenerative diseases, cancer, and even COVID-19. DNA methylation has proved informative for cfDNA deconvolution since up to a fifth of autosomal CpG sites vary by cell type. However, this deconvolution has proven to be no easy task. Thankfully, Noah Zeitlen’s group at UCLA has created CelFiE to apply a clarifying filter to low coverage whole-genome bisulfite sequencing (WGBS) cfDNA snapshots. Here are the perks of this new tool that these bioinformatician mediums created to commune with the dead (cells):
- CelFiE doesn’t rely on predefined sites: it determines a set of tissue informative markers (TIMs) on the fly by calculating the top 100 informative CpGs for each cell type in a reference set
- The expectation maximization (EM) approach allows for the possibility of unknown cell types comprising part of the cfDNA mixture
- This incorporation of uncertainty into CelFiE’s deconvolution logic prevents the inflation of cell-fraction estimates for other cell types
During their first cfDNA seance, the team validated CelFiE using WGBS data for:
- Pregnant vs non-pregnant women, where it correctly estimated a higher fraction of placental cfDNA in pregnant women, as well as a higher fraction for women in their second trimester than women in their first.
- ALS patients and controls from the University of Queensland, finding more skeletal muscle cfDNA in patients, with a higher level of significance than comparable methods.
- They validated this result using a cohort from UCSF
In general, quite a high proportion of cfDNA in both pregnant women and ALS patients was assigned an uncertain provenance. This indicates that using pre-selected reference CpG panels can preclude identifying novel biomarkers in conditions detectable using cfDNA. Ultimately, CeFiE’s selfies should prove useful for biomarker discovery and disease monitoring.
Learn how to see dead cells in Nature Communications, May 2021.