Overall, ARG inference tools have demonstrated remarkable robustness to computational phasing errors. While various practical challenges may limit ARG inference quality, we believe that computational phasing inaccuracies should not be a problem of concern.
7/7

Under a selective sweep model, we saw little difference in branch-length-based diversity between ARGs inferred from true and computationally phased haplotypes, demonstrating the practicality of applying ARGs inferred from unphased data in evolutionary analyses.
6/7

Under the CEU bottleneck model, we observed consistently minor effects of phasing errors. The difference between ARG inference methods is much larger than the difference in performance between using true haplotypes and computationally phased haplotypes.
5/7

In constant population size models, we observed only a slight reduction in ARG inference accuracy in terms of estimates of coalescence times or recombination breakpoint counts when using computationally phased haplotypes. The effect also reduced as the sample size increased.
4/7

Here, we show that ARG inference remains robust to phasing errors even at incredibly small sample sizes (8 haplotypes). We simulated ARGs and VCFs under various demographic models, simulated phasing errors, and compared ARGs inferred from true and computationally phased haplotypes.
3/7

ARGs are powerful tools in population genetics. However, ARG inference generally requires phased haplotypes. Phasing quality is limited at small sample sizes, making it difficult in non-model organisms due to the usually limited sample size and a lack of reference panels.
2/7

Overall, ARG inference tools have demonstrated remarkable robustness to computational phasing errors. While various practical challenges may limit ARG inference quality, we believe that computational phasing inaccuracies should not be a problem of concern.
7/7

Under a selective sweep model, we saw little difference in branch-length-based diversity between ARGs inferred from true and computationally phased haplotypes, demonstrating the practicality of applying ARGs inferred from unphased data in evolutionary analyses.
6/7

Under the CEU bottleneck model, we observed consistently minor effects of phasing errors. The difference between ARG inference methods is much larger than the difference in performance between using true haplotypes and computationally phased haplotypes.
5/7

In constant population size models, we observed only a slight reduction in ARG inference accuracy in terms of estimates of coalescence times or recombination breakpoint counts when using computationally phased haplotypes. The effect also reduced as the sample size increased.
4/7

Here, we show that ARG inference remains robust to phasing errors even at incredibly small sample sizes (8 haplotypes). We simulated ARGs and VCFs under various demographic models, simulated phasing errors, and compared ARGs inferred from true and computationally phased haplotypes.
3/7

ARGs are powerful tools in population genetics. However, ARG inference generally requires phased haplotypes. Phasing quality is limited at small sample sizes, making it difficult in non-model organisms due to the usually limited sample size and a lack of reference panels.
2/7