Orchid can identify embryos less likely to develop disease even after accounting for sibling attenuation

Orchid can identify embryos less likely to develop disease even after accounting for sibling attenuation
One way to lower the chance that a future child might develop a disease later in life is to use a genetic test to identify embryos with a lower risk for that disease. For chronic diseases like heart disease or schizophrenia, a single gene doesn’t usually drive genetic risk - rather hundreds to over one million DNA differences (variants) are aggregated into a genetic risk score that measures risk. A limitation of  genetic risk scores is sibling attenuation–they work best in unrelated people and are less predictive in related people like siblings (and therefore embryos). We have found that even after accounting for sibling attenuation, genetic risk scores can identify which adult siblings are more likely to develop a disease. Since genetic risk scores can identify adult siblings more likely to develop a disease, it makes sense they can do the same with embryos because embryos are siblings. Using genetic screening to identify embryos with higher risk can help parents have children less likely to develop disease later in life.
Written by Orchid Team

Key Points:

  • Genetic risk scores are less predictive between siblings (and embryos) compared to unrelated people. This is called sibling attenuation.  
  • Genetic risk scores can identify which adult sibling is less likely to develop a disease. This means even after accounting for the lower performance due to sibling attenuation, genetic risk scores still distinguish risk between siblings. 
  • Since genetic risk scores can identify adult siblings more likely to develop a disease, it makes sense they can do the same with embryos because embryos are siblings. 
  • Using genetic screening with embryos can help parents have children less likely to develop disease later in life. 

Orchid’s genetic risk scores can identify embryos less likely to develop a disease

Orchid’s mission is to help everyone have a healthy baby. We do this through genetic screening of embryos created during in vitro fertilization (IVF). A key part of this screening is identifying the embryo(s) with the lowest chance for developing any of a wide variety of diseases.  

For example, Orchid offers a genetic test that can determine the chance that an embryo will develop any of twelve chronic diseases that affect large numbers of people later in life. These diseases include Alzheimer’s disease, atrial fibrillation, bipolar disorder, breast cancer, celiac disease, coronary artery disease, inflammatory bowel disease, prostate cancer, severe obesity, schizophrenia, type 1 diabetes, and type 2 diabetes. This type of screening uses genetic risk scores made up of hundreds to over one million different DNA variants. 

Genetic risk scores are developed using large databases made up of unrelated people. This results in genetic risk scores working best in unrelated people and being less predictive in related people like siblings (and embryos)  [1,2,3]. This limitation goes by the name of sibling attenuation. (It is worth noting that identifying rare DNA variants and other larger DNA changes that can cause disease is unaffected by sibling attenuation.)

Because the predictive power of genetic risk scores can be reduced between siblings (and therefore embryos), it is important to measure how much sibling attenuation affects performance. We’ve found in datasets of adult siblings that genetic risk scores can distinguish the sibling who is more likely to develop disease from those who are less likely. Since genetic risk scores can identify adult siblings more likely to develop a disease, it makes sense they can do the same with embryos because embryos are siblings.  

Our results are consistent with two recent studies that showed the effects of sibling attenuation are modest for more physical traits like the diseases we test for [4,5]. Using genetic risk scores with embryos can help parents have children less likely to develop disease later in life. 

Genetic risk scores can identify which sister is more likely to develop breast cancer

For the following two studies, Orchid scientists determined what a typical difference in genetic risk scores would be between the lowest and the highest breast cancer genetic risk scores among five embryos. Five was chosen because that is a typical number of embryos available for screening with IVF.

Then we identified pairs of sisters with that difference in genetic risk scores in the UK Biobank and ran the following two analyses using those pairs of sisters: 

Analysis 1: Nearly three times as many high risk sisters developed breast cancer

In the first analysis, we looked amongst pairs of sisters where:

  • one sister had a higher breast cancer genetic risk score
  • one sister had a lower breast cancer genetic risk score
  • one of the sisters developed breast cancer

We found 170 pairs of sisters where this was the case. The numbers of sisters in each group who developed breast cancer are graphed below:

Figure 1: Breast cancer cases in high risk and low risk sisters. A total of 170 sister pairs were identified in the UK Biobank where two sisters had a typical difference in breast cancer genetic risk scores that would be reflected between the lowest and the highest risk scores among five embryos and one sister developed breast cancer. The dotted line indicates expected numbers if the genetic risk score was not predictive of breast cancer risk.

If the genetic risk scores weren’t predictive, around 85 of each risk group would develop breast cancer. Instead, 127 sisters with the higher genetic risk score developed breast cancer while only 43 sisters with the lower genetic risk score did. 

We would not expect to see any difference between the sisters in each pair if sibling attenuation had too large an effect on the predictive powers of genetic risk scores. Since we do see a difference, genetic risk scores can be used to identify the sibling (or embryo) less likely to develop breast cancer.

Analysis 2: Higher risk sisters are around 2.5 times more likely to develop breast cancer

In the second analysis, we calculated the percent of sisters who developed cancer at various ages for:

  • sisters with the higher genetic risk score
  • sisters with the lower genetic risk score
  • sisters where the genetic risk score was not taken into account (“random sisters”)

The data are graphed below:

Figure 2: Breast cancer risk in women across age for high risk siblings, low risk siblings, and randomly selected siblings. All sister pairs in the UK biobank were selected and the variables “year of birth”, “age at cancer diagnosis”, breast cancer status as inferred from self-reported and ICD10 diagnosis fields, were used to compute age- and group-stratified cumulative breast cancer prevalence rates for three groups: random, high risk, and low risk siblings. Since the prevalence rates displayed here have been corrected for age at diagnosis, they may differ from uncorrected prevalence rates reported elsewhere.

Around 15% of the sisters with the higher genetic risk score developed breast cancer by age 84 compared to only 6% of the sisters with the lowest genetic risk score. Again, we would not expect to see any difference between these two groups if sibling attenuation had too large an effect (if they erased all performance of the genetic risk score). Since we do see a difference, genetic risk scores can be used to identify the sibling (or embryo) less likely to develop breast cancer.

In addition, the sisters with the lowest genetic risk score developed breast cancer less often than the sisters for whom genetic risk scores were not taken into account (which mirrors untested embryos). These data indicate that genetic screening of embryos can provide real benefits over no testing at all even after accounting for sibling attenuation. 

How sibling attenuation affects accuracy of Orchid’s genetic risk scores

Studies have found that sibling attenuation can affect the accuracy of genetic risk scores by up to 50%. [1,2,3]. Orchid has found that their predictions among sibling pairs are 65% to 82% as accurate as predictions among pairs of unrelated people. The figure below shows the results for four Orchid genetic risk models:

Figure 3: Accuracy in sibling pairs as a percentage of accuracy among unrelated pairs. Within each group, accuracy is the correlation between the score difference and condition status difference among pairs (either sibling pairs, or pairs of unrelated people). Numbers of White British sibling pairs in which one sibling developed disease and the other did not used in this analysis for each disease: breast cancer, 1694; atrial fibrillation, 1799; coronary artery disease, 1856; type 2 diabetes,: 2256.

While genetic risk scores are less predictive between related people, they are still powerful enough to identify the sibling (and therefore the embryo) with a lower chance for developing a disease. Taken together, the results presented here indicate that genetic screening of embryos can help parents have children less likely to develop disease later in life. 

In an IVF environment where only some embryos will be implanted, genetic risk scores can provide valuable information to prospective parents in deciding which embryo(s) to implant. In other words, genetic screening of embryos using genetic risk scores can provide a better outcome than without screening, when the embryo implanted is chosen at random. 

References 

  1. Selzam S, Ritchie SJ, Pingault JB,  Reynolds CA,.O’Reilly PF, Plomin R. Comparing Within- and Between-Family Polygenic Score Prediction. Am J Hum Genet 2019; 105: 351-363. https://www.cell.com/ajhg/fulltext/S0002-9297(19)30231-9  
  2. Kong A, Thorleifsson G, Fridge ML, Vilhjalmsson BJ, Youngthorgeir AI, Thorgeirsson TE, et al. The nature of nurture: Effects of parental genotypes. Science 2018; 359: 424-428 https://www.science.org/doi/10.1126/science.aan6877 
  3. Turley P, Meyer MN, Wang N, Cesarini D, Hammonds E, Martin AR, et al. Problems with Using Polygenic Scores to Select Embryos. N Engl J Med 2021; 385: 78-86 https://www.nejm.org/doi/full/10.1056/NEJMsr2105065 
  4. Lello L, Raben TG, Hsu SDH. Sibling validation of polygenic risk scores and complex trait prediction. Scientific Reports 2020; 10: 13190 https://doi.org/10.1038/s41598-020-69927-7 
  5. Howe LJ, Nivard MG, Morris TT, Hansen AF, Rasheed H, et. al. Within-sibship GWAS improve estimates of direct genetic effects. bioRxiv 2021;  https://doi.org/10.1101/2021.03.05.433935 
get access

Get expert reviewed guides hot off the presses.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Recent Articles