Skip to main content
Premium Trial:

Request an Annual Quote

Ebola Survivors Have Different Transcriptome Profiles From Those Who Die

NEW YORK (GenomeWeb) – Transcriptomic profiles can distinguish people with Ebola who are more likely to survive from those who are more likely to die, according to a new study.

An international team of researchers examined the transcriptomes of blood samples taken from people infected with Ebola and from others recovering from the disease. The patients were infected during the West African Ebola outbreak in 2014 and 2015.

As they reported in Genome Biology, Julian Hiscox at the UK's National Institute of Health Research and his colleagues found that people who eventually died of Ebola had higher levels of interferon signaling and acute phase responses, as compared to those who later survived. Through this, the researchers developed a panel of genes whose expression could predict patient outcome independently of viral load.

"Integration of this type of analysis in future outbreak responses could help direct therapy and maximize a beneficial outcome of infection," the authors wrote, "particularly in the development of diagnostic approaches that can accurately stratify patients for treatment based on the likely outcome of infection."

He and his colleagues deeply sequenced blood samples obtained from 138 patients in Guinea, which had been primarily collected for qRT-PCR-based diagnosis of Ebola. Twenty-six samples had to be discarded, leaving 112 samples, 24 of which came from patients who survived and 88 of which came from patients who died. None of these patients received intensive palliative or experimental care.

For a baseline, the researchers also sequenced samples from 16 survivors after they'd recovered from their infection and who were negative for the virus by PCR and from six healthy volunteers in British Columbia who were not exposed to the virus.

Using that data, the researchers uncovered some 1,300 genes that were more highly expressed in the survivor group as compared to the recovered group and 2,200 genes that were more highly expressed in the fatal group as compared to the recovered group, though about half of those overlapped with the survivor group.

Blood samples from people with acute infections with either ultimate outcome exhibited increased abundance of pro-inflammatory factors, including CXCL10, CCL2/MCP-1, CCL8/MCP2, and CXCL11. Gene-pathway analysis likewise noted that infected people had an enrichment of genes within the same signaling pathways, particularly in interferon signaling, complement, and coagulation pathways. A similar pattern of enrichment was also found through Ingenuity Pathway analysis.

Despite these similarities, there were differences in transcriptome patterns between acute-stage patients who ultimately survived and those who ultimately died. Some 246 transcripts were differentially expressed, most of which were increased among fatal cases. This differential abundance was most striking among genes associated with coagulation and acute phase signaling, the researchers noted.

Hiscox and his colleagues turned to independent machine learning-based methods to develop predictive models based on these differentially expressed genes. With a support vector machine-approach using the top 10 genes, they developed a model with 79 percent accuracy — better, they noted, than gauging outcomes by viral load.

A substitution method on groups of 10 randomly selected genes performed even better, with accuracy ranging from 79 percent to 85 percent; a random forest approach yielded an 89 percent accuracy; and a paired-gene profiling method yielded 92 percent accuracy.

Common genes to each set were TGFB1 — which encodes an extracellular matrix protein — VACM1 — which encodes protein involved in lymphocyte extravasation — and HOPX.

They further tested the classifiers on a set of 20 patients whose viral load fell in a gray zone where outcomes can't be determined to find that they could predict outcomes, independent of viral load. This suggested to them that their biomarkers could be used during an outbreak to help allocate resources.

"The ability to triage patients by disease severity and likely outcome can be of practical benefit for patient care," the researchers wrote.