Population divergence and connectivity through time.
To estimate population divergence and connectivity through time, we used
the program MSMC-IM (Wang et al., 2020). MSMC-IM uses output from MSMC
(Schiffels & Durbin, 2014) to fit an isolation with migration model to
coalescent rates estimated within and among two populations (Wang et
al., 2020). We performed MSMC-IM analyses separately for each species.
We initially tried to run MSMC with all sampled individuals, but we were
unable to get our runs to complete due to computational limitations. We
therefore chose two random individuals per species per population (N = 4
total for each species except the Abyssinian Catbird where N = 3). Here,
we used WhatsHap (Martin et al., 2016) to phase genotypes for all
individuals. We chose this method because it uses read-based phasing and
does not require large sample sizes or reference SNP panels for phasing.
Phasing is required for these analyses because MSMC-IM requires
estimating cross-coalescence rates between at least two individuals
(i.e., ≥ four chromosomes). This is in contrast to MSMC demographic
analyses with a single individual that do not require phasing between
the two chromosomes sampled in a single diploid individual. We masked
all regions with sequencing coverage lower that six in any individual to
minimize inclusion of sites with phasing or genotyping errors. We ran
MSMC with up to 20 iterations and 23 distinct time segments, as with
individual-based demographic histories. Recent within-population
histories were qualitatively similar to those estimated from single
individuals but with less temporal resolution (results not shown). Less
resolution may be expected since we are masking any genomic regions with
low sequencing coverage in any individual, and therefore sampling less
of the genome. Regardless, our main goal here was to estimate whether
all our focal species shared the same patterns of population divergence
and migration through time, and small shifts in overall resolution would
not heavily impact these types of interpretations. For each species, we
used the MSMC output as input for MSMC-IM to fit an isolation with
migration model. In MSMC-IM, we used estimated empirical mutation rates
for each species, and the recommended regularization settings for
estimating migration rates and population sizes (1e-8 and 1e-6,
respectively).