Skip to Main Content

While the world waited for He Jiankui to back up his claim of CRISPR’ing two babies to an international summit of genome editors in Hong Kong on Wednesday morning, STAT asked genomics experts to review the only data He has publicly revealed, in a spreadsheet on the Chinese Clinical Trials Registry. Their verdict: The quality of the data is low, and there’s too little of it, to draw any firm conclusions about what He’s team accomplished.

And there is no proof that the two embryos described in the spreadsheet were used to produce the twins whose birth He announced earlier this week.


An Excel spreadsheet linked to the listing for He’s trial includes data from scanning the genomes of two embryos — labeled as “Embryo #1” and “Embryo #2.” The tables for each tally single-nucleotide variants — differences in a single A, T, C, or G in the DNA sequence in a specific spot on the genome — and characteristics that may show up in someone who has those variants.

Unlike in ClinVAR, a free public archive based on reports of human variations and how they show up in people, there is no sign of interpretation of what those variants might mean. Some of the variants in He’s tables are classified as both “likely pathogenic” and “benign” for rare conditions such as sporadic abdominal aortic aneurysm, for example. None of them says whether a variant is dominant or recessive.

The tables don’t comport with scientific or clinical standards, the experts said, making them leery of judging their importance.


“I would say it is not professional practice to put tables of variant interpretations directly out of ClinVAR without added evidence review and a clinical reporting context of what this means for the patient,” said Heidi Rehm, medical director of the Broad Institute’s Clinical Research Sequencing Platform. She is also chief genomics officer at Massachusetts General Hospital.

“It’s a pretty minimal report,” said David Liu, a genome editing expert at Harvard University.

The report describes some of the variants as “de novo” mutations, which are not present in the parents, and thus are presumed to have occurred in the sperm, egg, or developing embryo, either through common errors in the copying of DNA that occurs when cells divide or through environmental influences such as cosmic rays or toxic chemicals.

“It’s also possible that these mutations that were scored as ‘de novo’ may have resulted from the procedure used by He’s group,” Liu said. “There is insufficient information to tell based only on this sparse report.”

The genetic analysis was reportedly performed when the embryos were a few days old by removing a single cell, much as is done for standard preimplantation genetic diagnosis when there is a concern that one or more of the embryos created by IVF might carry a dangerous mutation; PGD allows parents to choose only healthy embryos to implant. Additional genetic sequencing and analysis was performed when the embryos were slightly older, and again at birth; those data have not been reported in the Chinese Clinical Trials Registry.

As a result, a key question remains unanswered: Did the genome editing work in most or even all cells, or only in some? If the latter, which is called mosaicism, it is possible that the girls do not have full protection against HIV infection. He’s aim was to edit a gene called CCR5 to give the babies immunity to HIV infection throughout their lives.

Dennis Grishin, co-founder of the startup DNA sequencing company Nebula Genomics, pointed out that both embryos have the same mutation on chromosome 19. “This might be a CRISPR off-target site that is replicated in both embryos,” he said. (CRISPR can inadvertently cut DNA in the wrong place, called an off-target site.)

“One should look at the sequence at this site and check if it has any similarity with” the CRISPR molecules used to edit the CCR5 gene, Grishin said. “On the other hand, it does not look like a CRISPR-induced mutation,” he added, “since those are typically deletions and insertions rather that single-base substitutions,” as this mutation is.

The main takeaway from the report is that embryo #2 got both copies of CCR5 edited, while embryo #1 has only one edited copy, Grishin said.

Rehm did put some stock in one of the spreadsheet’s columns. Labeled “Evidence Level,” all the entries say “Low” or “0.”