Search for: All records

Reproductive isolation is necessary for population divergence to lead to the formation of separate species. This can occur due to physical isolation of populations, which drives allopatric speciation, or other methods of isolation, such as sympatric speciation where the diverging species are physically in the same range, but structural genomic changes or mutations cause the population to diverge into two different species. Parapatric speciation occurs when populations that are geographically adjacent to each other diverge, which can be driven by adaptations to environmental differences, even with ongoing gene flow. Two desert- adapted brittlebush species, Encelia farinosa and Encelia californica, diverged less than 1 million years ago (Singhal et al., 2020) and have a parapatric distribution, residing in different environments in the Mojave and Sonoran deserts. Encelia farinosa (Brittlebush) has unique silvery leaves that are covered in tiny hairs (leaf pubescence) to better control leaf temperature in the hot and arid conditions of the Sonoran Desert. Encelia californica (California Brittlebush) does not display leaf pubescence and is found in a smaller region of the Mediterranean-like environment of the west coast of North America. Encelia californica is exposed to more precipitation than most other Encelia species. Even with their different morphologies, these two species are still able to hybridize and create fertile offspring (Clark, 1998). Using PacBio sequencing and Hi-C scaffolding, we assembled and annotated reference genomes for both species to investigate the genomic basis of reproductive isolation in these two species. The scaffold N50/L50 are 10 scaffolds and 76.3 Mbp, and 12 scaffolds and 64.5 Mbp for E. farinosa and E. californica, respectively. Using comparative genomic analyses such as tests for differential adaptation and chromosomal translocations will help reveal whether the drivers of speciation in the Encelia radiation were external (e.g., geologic/climatic) or internal (e.g., genomic rearrangement). These analyses will also help answer how accumulated genomic differences can cause speciation in populations that are not geographically isolated. Analyses such as these are new, exciting sources of information for testing geogenomic and other Earth- life hypotheses. 

Understanding the timescales on which different geologic processes influence genetic divergence is crucial to defining and testing geogenomic hypotheses and characterizing Earth- life evolution. To see if we can recover a genetic signal produced by a hypothetical physical barrier to gene flow, we used a geographically explicit simulation approach. We used the CDMetaPop software to simulate heritable genetic, nonadaptive, data for 20 geographically distinct populations distributed throughout the Baja California peninsula of Mexico, a landscape where a transpeninsular seaway barrier has been proposed to have isolated the southern peninsula and caused the observed latitudinal genetic divergence in over 80 terrestrial species. We simulated 10,000 generations of isolation by a barrier under two dispersal scenarios (1 km and 100 km of max. dispersal from population of origin per generation) and three DNA substitution rates (10-7, 10-8 and 10-9 nucleotide substitutions per site per generation). Our simulations indicate that a physical barrier can produce strong genetic divergence within 10,000 generations, comparable to the continuum of values observed in nature for different taxonomic groups and geological settings. We found that the generation time of the organism was by far the most important factor dictating the rate of divergence. Evaluating different generation times (0.02, 0.2, 2 and 20 years), showed that species with longer generation times require longer periods of isolation to accumulate genetic divergence over 10k generations (~1 My). Simulating 10,000 generations of gene flow following removal of the barrier showed that the divergence signal eroded quickly, in less than 1,000 generations in every scenario, a pattern supported by theory from population genetics. These results are particularly relevant to geogenomic studies because they show that ephemeral gene flow barriers produce different magnitudes of genetic signals depending on attributes of the organism, particularly generation time, and that if reproductive isolation is not achieved during isolation, then the evolutionary signal of an ephemeral barrier may not develop. This work helps guide the limits of detectability when integrating genomic data with geological and climatic processes. 

Central Baja California (BC) experienced tectonism and volcanism that shaped the landscape from the Miocene to Recent. One important feature is the San Ignacio trough (SIT) that hosted a marine seaway or embayment and acted as a physical barrier to animal and plant migration. This barrier may be responsible for a well-known break in the DNA, N and S of this region. Central BC has also hosted contemporary voluminous and chemically diverse volcanism. Radiometric ages provide important constraints on the origins and longevity of critical topographic features. The Baja GeoGenomics research group is investigating the nature and timing of Pliocene marine and tidal deposits in the NE-oriented, low-lying SIT, located W of the peninsular divide. These new data reveal that the Sierra San Francisco, a highland volcanic area immediately N of the SIT, is a series of volcanoes constructed of dacitic and andesitic Peleean domes with voluminous lahar and pyroclastic flow deposits. These calcalkaline rocks were previously thought to be subduction-related magmatism and part of the early to middle Miocene (~2412 Ma) Comondu Group. However, zircon U-Pb and 40Ar/39Ar dates yield ages of 11-9 Ma. These data indicate the Sierra San Francisco erupted post-subduction and is not part of the lithologically similar but older Comondu Group. Within the SIT, 12km NE of San Ignacio at 200 m asl, newly mapped marine tidal deposits, informally called the San Regis beds, indicate that the SIT has been significantly uplifted. Mafic scoria interbedded in tidal deposits yield a groundmass 40Ar/39Ar age of about 4.2 0.1 Ma. San Regis tidal beds are unconformably overlain by a rhyolite ash-flow tuff from the Quaternary La Reforma caldera situated to the E, on the Gulf of California coast. The highly mobile ash cloud flowed W into the SIT at least as far as the San Regis beds locality NE of San Ignacio. The tuff yielded a preliminary U-Pb zircon age of 1.09 0.04 Ma and an 40Ar/39Ar anorthoclase age of 1.11± 0.01 Ma. These dates indicate that the ash-flow was one of the latest erupted from the caldera and its distribution was in part controlled by the SIT. In BC genetic diversity along the peninsula appears to change at the latitude of the SIT. Tidal and volcanic deposits suggest this topographic low persisted for over 4Ma and remains a distinctive feature in the topography today. 

Late Cenozoic evolution of the Baja California (BC) peninsula governs its species diversity, with changes to terrestrial habitats and shorelines driven by volcanic and tectonic processes. New geologic mapping and geochronology in central BC help assess if recent landscape evolution created a barrier to gene flow. The NW-trending topographic divide of the BC peninsula near San Ignacio-Santa Rosalia (27.4N) is a low (400500 m asl), broad (2030 km-wide) pass. At the pass, ~2022-Ma volcaniclastic strata, mafic lavas, fluvial conglomerate, cross-bedded eolian sandstone, and a felsic tuff dip ~515 SW. Similar lithology and chronology suggest these strata correlate to the lower Comondu Group (CG). They are overlain by middle Miocene (~1114 Ma) mafic lavas with similar SW dips that overlap in age with the upper CG. NW of the pass, upper Miocene (~9.511 Ma) post-CG volcaniclastic strata and mafic lava flows are exposed in the Sierra San Francisco and dip ~10 SE on its SE flank, inclined differently than older SW-dipping CG at the pass. The basalt of Esperanza (~10 Ma) unconformably overlies the CG at and west of the pass. Its ~1 regional dip suggests that ~515 of SW tilting occurred prior to ~10 Ma in the footwall of the NW-striking Campamento fault, located at the base of the ~150 m-high rift escarpment. The N-striking Arroyo Yaqui fault, ~10 km E of the Campamento fault in a low-relief region capped by Quaternary marine strata, exposes crystalline basement in its footwall and may be a major rift margin structure. Thus the location, orientation, and age of the divide may be controlled by rift-related faulting and tilting plus beveling and lateral retreat of the escarpment. Pliocene tidal sediments occur up to ~200 m asl ~20 km west of the low pass similar to Pliocene marine strata east of the pass at ~300 m asl, indicating late Miocene to Pliocene subsidence was followed by >200 m of post-4 Ma uplift. Uplift was likely driven by transtensional faulting and possibly magmatic inflation by ~7090 km-wavelength domes. Further mapping will constrain the timing of vertical crustal motions and test whether the tidal embayment crossed the peninsula through this low pass, isolated species, and prevented terrestrial gene flow. Integration of geologic and genetic data will determine how volcano- tectonic processes shaped genetic diversity. 

ubmodular functions have applications throughout machine learning, but in many settings, we do not have direct access to the underlying function f . We focus on stochastic functions that are given as an expectation of functions over a distribution P. In practice, we often have only a limited set of samples fi from P . The standard approach indirectly optimizes f by maximizing the sum of fi. However, this ignores generalization to the true (unknown) distribution. In this paper, we achieve better performance on the actual underlying function f by directly optimizing a combination of bias and variance. Algorith- mically, we accomplish this by showing how to carry out distributionally robust optimiza- tion (DRO) for submodular functions, pro- viding efficient algorithms backed by theoret- ical guarantees which leverage several novel contributions to the general theory of DRO. We also show compelling empirical evidence that DRO improves generalization to the un- known stochastic submodular function.