Search for: All records

T. Kim Dao, Kailey Ferger, J. David Lambert, A chitin-binding domain-containing gene is essential for shell development in the mollusc Tritia, Developmental Biology, Volume 520, 2025, Pages 1-12, ISSN 0012-1606, https://doi.org/10.1016/j.ydbio.2024.12.016. (https://www.sciencedirect.com/science/article/pii/S0012160624002884) Abstract: Mollusc shells are diverse in shape and size. They are created by a shell epithelium which secretes a chitinous periostracum membrane at the growing edge of the shell, and then coordinates biomineral deposition on the underside of this membrane. Although mollusc shells are important for studying the evolution of morphology, the molecular basis of the shell development is poorly understood. In this paper, we investigate genes involved in the shell development of the gastropod mollusc Tritia (previously known as Ilyanassa). We characterize the contributions of the 2d micromere to the shell and other non-shell structures. We identify eight shell-specific genes and five non-shell specific genes by comparing the transcriptomes of wild-type and 2d ablated embryos. Morpholino knockdown of one of the shell-specific genes, ToChitin-binding domain-containing (ToChitin BD), results in shell defects. The chitinous periostracal membranes in ToChitin BD morpholino knockdown embryos lose their well-defined edge and peroxidase gradient. 

The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15.5B parameter models with 8K context length, infilling capabilities and fast large-batch inference enabled by multi-query attention. StarCoderBase is trained on 1 trillion tokens sourced from The Stack, a large collection of permissively licensed GitHub repositories with inspection tools and an opt-out process. We fine-tuned StarCoderBase on 35B Python tokens, resulting in the creation of StarCoder. We perform the most comprehensive evaluation of Code LLMs to date and show that StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40% pass@1 on HumanEval, and still retains its performance on other programming languages. We take several important steps towards a safe open-access model release, including an improved PII redaction pipeline and a novel attribution tracing tool, and make the StarCoder models publicly available under a more commercially viable version of the Open Responsible AI Model license. 

Abstract Predicting the daily variability of Equatorial Plasma Bubbles (EPBs) is an ongoing scientific challenge. Various methods for predicting EPBs have been developed, however, the research community is yet to scrutinize the methods for evaluating and comparing these prediction models/techniques. In this study, 12 months of co‐located GPS and UHF scintillation observations spanning South America, Atlantic/Western Africa, Southeast Asia, and Pacific sectors are used to evaluate the Generalized Rayleigh‐Taylor (R‐T) growth rates calculated from the Thermosphere Ionosphere Electrodynamics General Circulation Model (TIEGCM). Various assessment metrics are explored, including the use of significance testing on skill scores for threshold selection. The sensitivity of these skill scores to data set type (i.e., GPS versus UHF) and data set size (30, 50, 60, and 90 days/events) is also investigated. It is shown that between 50 and 90 days is required to achieve a statistically significant skill score. Methods for conducting model‐model comparisons are also explored, including the use of model “sufficiency.” However, it is shown that the results of model‐model comparisons must be carefully interpreted and can be heavily dependent on the data set used. It is also demonstrated that the observation data set must exhibit an appropriate level of daily EPB variability in order to assess the true strength of a given model/technique. Other limitations and considerations on assessment metrics and future challenges for EPB prediction studies are also discussed.