Model embryos created using tessera with 20% aneuploid cells (dark grey) and 80% euploid cells (light green), shown from front and back. Embryos have either high (A) or low (B) dispersal of the aneuploid cells. The potential biopsies obtained from each differ (graph C for embryo A, graph D for embryo B). For the clustered embryo (B) most biopsies are not representative of the mosaicism level present in the embryo (D); the reverse is true for the dispersed embryo.

Difference between the aneuploidy of a 5-cell biopsy and the aneuploidy of the embryo for embryos with increasing levels of (A) aneuploidy at constant dispersal of 1 and (B) increasing dispersal with constant aneuploidy of 20%. Values show the mean and standard deviation of 100 replicates. Error is highest at 40%-60% aneuploidy, and at low dispersals. The local minima in the aneuploidy chart are due to the 5-cell biopsy size, allowing exact matches to the biopsy at these aneuploidies.

The percentage of biopsies in the same class as the embryo from which they came, for all combinations of aneuploidy and dispersal. Each combination represents the mean of 100 embryos. Accuracy is lower in embryos with low dispersal, especially in ‘low-level’ mosaic embryos.

The effect of increasing biopsy size on accuracy for all combinations of embryo aneuploidy and dispersal. Numbers above each heatmap show the biopsy size. There are only small increases to accuracy over the 5-10 cell biopsy range, and even at larger biopsy sizes there is little difference to accuracy when dispersal of aneuploid cells is very low.

The possible origin embryos for 5-cell biopsies with 0 to 5 aneuploid cells. Biopsies with either 0 or 5 aneuploid cells are more likely to come from a constitutively normal or constitutively aneuploid embryo respectively, but mosaic biopsies can come from a much broader possible range of embryos.

A mosaic biopsy has limited predictive power at classifying an individual embryo. (A) The percentage of biopsies with a given aneuploidy originating from an embryo with a given aneuploidy at three levels of dispersal (0, 0.5, 1). The embryo classification thresholds are drawn as white squares; for prediction to be useful, the majority of embryo aneuploidies should be within the squares. (B) the percentage of biopsies that correctly predict their embryo class at the three levels of dispersal.

Effect of the size of aneuploidy differences on ranking two embryos with different levels of aneuploidies by biopsy result for embryos with zero dispersal. A) There is a greater than 50% chance of unambiguously choosing the correct rank order as long as the absolute difference in aneuploidy is greater than 25%. B) Biopsies with equal aneuploidy (tied ranking) are evenly split between correct and incorrect ranking to mimic the clinical scenario in which a random choice will be correct 50% of the time. Ranking two embryos based on biopsy outcomes will always on average correctly rank embryos better than chance if the embryos have unequal levels of aneuploidy. 100 replicate embryos were generated per aneuploidy combination. Values show mean and standard deviation after aggregating by aneuploidy difference. More detailed breakdowns by aneuploidy and dispersal combinations are shown in supplementary data (Figures S4 - S5).

Single biopsies can reliably select the best three embryos from a pool of six embryos. For each dispersal level, 100 pools of embryos were generated with random aneuploidy levels. A single biopsy from each embryo was used to rank the pool, and the three embryos with lowest rank were selected. The figure shows the mean percentage of selected embryos that are in the true “best three”. Values show mean and standard deviation from 100 embryo pools. Full combinations are in Figure S6.

Implantation (light grey) and ongoing pregnancy/birth rates (dark grey) split by segmental or whole chromosome aneuploidies. Complex mosaics with multiple changes detected were classified based on the largest abnormality present. Thus ‘Whole chromosome mosaics’ includes all cases involving a full chromosome aneuploidy regardless of the presence or absence of additional segmental aneuploidies, while ‘Segmental mosaics’ includes only cases where all aneuploid regions were sub-chromosomal in extent. Breaking the data into <50% or ≥50% aneuploidy reveals significant differences in outcome for whole chromosome mosaics, though not for segmental mosaics (chi square tests with Bonferroni correction). This figure is structured to match Viotti et al. 2021 Figure 2A for comparison.

Accuracy of one 10-cell biopsy across embryo sizes and dispersals. Compare to Figure 6 (one 5-cell biopsy) and Figures S2 and S3 (two 5-cell biopsies).

Percentage of two 5-cell biopsies matching their embryo class embryo sizes and dispersals. Compare to the single 5-cell biopsy in Figure 6B and the single 10-cell biopsy in Figure S1.

Comparison of one 10-cell biopsy and two 5-cell biopsies. Accuracy is increased with two 5-cell biopsies for embryos with low dispersal by up to ∼12 percentage points. There are no benefits at higher dispersals. This chart is the difference between Figures S1 and S2.

Effect of dispersal and aneuploidy differences on ranking two embryos with different aneuploidies by biopsy result. There is a greater than 50% chance of choosing the correct rank order at all levels of dispersion as long as the absolute difference in aneuploidy is greater than 20%. 100 replicate embryos were generated per aneuploidy and dispersal combination. Values show mean and standard deviation after aggregating by aneuploidy difference. Top left panel is equivalent to Fig 7A.

Contrast with Fig S4. Effect of dispersal and aneuploidy differences on ranking two embryos with different aneuploidies by biopsy result. Biopsies with equal aneuploidy (tied ranking) have been evenly split between correct and incorrect ranking, since a random choice will be correct 50% of the time. Ranking two embryos based on biopsy outcomes will always on average rank embryos better than chance if the embryos have unequal levels of aneuploidy. Top left panel is equivalent to Fig 7B.

The ability to select the best k embryos from a pool of n when the pool contains embryos with random levels of aneuploidies. Embryos are correctly selected better than chance in almost all cases (the exception at large pool sizes and low selection sizes). All embryos in the pool had 200 cells. Values are mean and standard deviation of 100 replicates. This figure complements Figure 8.

Logistic regression on outcomes shows a significant association with aneuploidy level and with the type of chromosomal abnormality. Points show the individual embryos, and are jittered for clarity. Lines show predicted outcomes for a given aneuploidy level from the logistic regression models and the shaded areas show the standard error. Both aneuploidy type and aneuploidy level have significant contributions to outcomes.