Search for: All records

We introduce a new method for identifying genomic regions that are rich in introgressed DNA. The method does not search for introgressed haplotypes on individual chromosomes. For this reason, it is impervious to differences among populations in the length distributions of introgressed fragments. The method works with or without a genome from the donor population. When such a genome is available, however, the method has greater statistical power. We apply the method to four previously-documented episodes of admixture: (1) from Neanderthals into modern Eurasians, (2) from a distantly related "superarchaic" population into Denisovans, (3) from this same superarchaic into Neanderthal-Denisovan ancestors, and (4) from a different superarchaic population into the ancestors of modern humans. Our estimates confirm that Neanderthal admixture is generally low in regions that previous studies have identified as "deserts"--genomic regions lacking archaic admixture. However, our method also identifies "oases"--small regions of inflated admixture--within these deserts. For all four episodes of admixture, we show that selection has opposed admixture near exons and regulatory elements. 

For the past decade, temporal annotation has been sparse: only a small portion of event pairs in a text was annotated. We present NarrativeTime, the first timeline-based annotation framework that achieves full coverage of all possible TLINKs. To compare with the previous SOTA in dense temporal annotation, we perform full re-annotation of the classic TimeBankDense corpus (American English), which shows comparable agreement with a signigicant increase in density. We contribute TimeBankNT corpus (with each text fully annotated by two expert annotators), extensive annotation guidelines, open-source tools for annotation and conversion to TimeML format, and baseline results. 

Abstract The Chemistry and Mineralogy X-ray diffraction (XRD) instrument aboard the Curiosity rover consistently identifies amorphous material at Gale Crater, which is compositionally variable, but often includes elevated sulfur and iron, suggesting that amorphous ferric sulfate (AFS) may be present. Understanding how desiccating ferric sulfate brines affect the spectra of Martian material analogs is necessary for interpreting complex/realistic reaction assemblages. Visible and near-infrared reflectance (VNIR), mid-infrared attenuated total reflectance (MIR, FTIR-ATR), and Raman spectra, along with XRD data are presented for basaltic glass, hematite, gypsum, nontronite, and magnesite, each at three grain sizes (<25, 25–63, and 63–180μm), mixed with ferric sulfate (+/−NaCl), deliquesced, then rapidly desiccated in 11% relative humidity or via vacuum. All desiccated products are partially or completely XRD amorphous; crystalline phases include starting materials and trace precipitates, leaving the bulk of the ferric sulfate in the amorphous fraction. Due to considerable spectral masking, AFS detectability is highly dependent on spectroscopic technique and minerals present. This has strong implications for remote and in situ observations of Martian samples that include an amorphous component. AFS is only identifiable in VNIR spectra for magnesite, nontronite, and gypsum samples; hematite and basaltic glass samples appear similar to pure materials. Sulfate features dominate Raman spectra for nontronite and basaltic glass samples; the analog material dominates Raman spectra of hematite and gypsum samples. MIR data are least affected by masking, but basaltic glass is almost undetectable in MIR spectra of those mixtures. NaCl produces similar FTIR-ATR and Raman features, regardless of analog material. 

The recent explosion in question answering research produced a wealth of both factoid reading comprehension (RC) and commonsense reasoning datasets. Combining them presents a different kind of task: deciding not simply whether information is present in the text, but also whether a confident guess could be made for the missing information. We present QuAIL, the first RC dataset to combine text-based, world knowledge and unanswerable questions, and to provide question type annotation that would enable diagnostics of the reasoning strategies by a given QA system. QuAIL contains 15K multi-choice questions for 800 texts in 4 domains. Crucially, it offers both general and text-specific questions, unlikely to be found in pretraining data. We show that QuAIL poses substantial challenges to the current state-of-the-art systems, with a 30% drop in accuracy compared to the most similar existing dataset.