Iconographic evidence from Egypt suggests that watermelon pulp was consumed there as a dessert by 4,360 BP. Earlier archaeobotanical evidence comes from seeds from Neolithic settlements in Libya, but whether these were watermelons with sweet pulp or other forms is unknown. We generated genome sequences from 6,000- and 3,300-year-old seeds from Libya and Sudan, and from worldwide herbarium collections made between 1824 and 2019, and analyzed these data together with resequenced genomes from important germplasm collections for a total of 131 accessions. Phylogenomic and population-genomic analyses reveal that (1) much of the nuclear genome of both ancient seeds is traceable to West African seed-use "egusi-type" watermelon (Citrullus mucosospermus) rather than domesticated pulp-use watermelon (Citrullus lanatus ssp. vulgaris); (2) the 6,000-year-old watermelon likely had bitter pulp and greenish-white flesh as today found in C. mucosospermus, given alleles in the bitterness regulators ClBT and in the red color marker LYCB; and (3) both ancient genomes showed admixture from C. mucosospermus, C. lanatus ssp. cordophanus, C. lanatus ssp. vulgaris, and even South African Citrullus amarus, and evident introgression between the Libyan seed (UMB-6) and populations of C. lanatus. An unexpected new insight is that Citrullus appears to have initially been collected or cultivated for its seeds, not its flesh, consistent with seed damage patterns induced by human teeth in the oldest Libyan material.

Motivation: Limited data access has hindered the field of precision medicine from exploring its full potential, e.g. concerning machine learning and privacy and data protection rules. Our study evaluates the efficacy of federated Random Forests (FRF) models, focusing particularly on the heterogeneity within and between datasets. We addressed three common challenges: (i) number of parties, (ii) sizes of datasets and (iii) imbalanced phenotypes, evaluated on five biomedical datasets. Results: The FRF outperformed the average local models and performed comparably to the data-centralized models trained on the entire data. With an increasing number of models and decreasing dataset size, the performance of local models decreases drastically. The FRF, however, do not decrease significantly. When combining datasets of different sizes, the FRF vastly improve compared to the average local models. We demonstrate that the FRF remain more robust and outperform the local models by analyzing different class-imbalances. Our results support that FRF overcome boundaries of clinical research and enables collaborations across institutes without violating privacy or legal regulations. Clinicians benefit from a vast collection of unbiased data aggregated from different geographic locations, demographics and other varying factors. They can build more generalizable models to make better clinical decisions, which will have relevance, especially for patients in rural areas and rare or geographically uncommon diseases, enabling personalized treatment. In combination with secure multi-party computation, federated learning has the power to revolutionize clinical practice by increasing the accuracy and robustness of healthcare AI and thus paving the way for precision medicine. Availability and implementation: The implementation of the federated random forests can be found at https://fea turecloud.ai/.

Hellweger et al. (Reports, 27 May 2022, pp. 1001) predict that phosphorus limitation will increase concentrations of cyanobacterial toxins in lakes. However, several molecular, physiological, and ecological mechanisms assumed in their models are poorly supported or contradicted by other studies. We conclude that their take-home message that phosphorus load reduction will make Lake Erie more toxic is seriously flawed.

Age-depth relationships are the key elements in paleoenvironmental studies to place proxy measurements into a temporal context. However, potential influencing factors of the available radiocarbon data and the associated modeling process can cause serious divergences of age-depth relationships from true chronologies, which is particularly challenging for paleolimnological studies in Arctic regions. This paper provides geoscientists with a tool-assisted approach to compare outputs from age-depth modeling systems and to strengthen the robustness of age-depth relationships. We primarily focused on the development of age determination data from a data collection of high-latitude lake systems (50 to 90 circle N, 55 sediment cores, and a total of 602 dating points). Our approach used five age-depth modeling systems (Bacon, Bchron, clam, hamstr, Undatable) that we linked through a multi-language Jupyter Notebook called LANDO ("Linked age and depth modeling"). Within LANDO we implemented a pipeline from data integration to model comparison to allow users to investigate the outputs of the modeling systems. In this paper, we focused on highlighting three different case studies: comparing multiple modeling systems for one sediment core with a continuously deposited succession of dating points (CS1), for one sediment core with scattered dating points (CS2), and for multiple sediment cores (CS3). For the first case study (CS1), we showed how we facilitate the output data from all modeling systems to create an ensemble age-depth model. In the special case of scattered dating points (CS2), we introduced an adapted method that uses independent proxy data to assess the performance of each modeling system in representing lithological changes. Based on this evaluation, we reproduced the characteristics of an existing age-depth model (Lake Ilirney, EN18208) without removing age determination data. For multiple sediment cores (CS3) we found that when considering the Pleistocene-Holocene transition, the main regime changes in sedimentation rates do not occur synchronously for all lakes. We linked this behavior to the uncertainty within the dating and modeling process, as well as the local variability in catchment settings affecting the accumulation rates of the sediment cores within the collection near the glacial-interglacial transition.

The growing number of wheat-related allergies worldwide has resulted in a new trend towards gluten-free alternatives. In this context, alternative cereals such as sorghum and oats are attracting new interest. Given the limited data available, the question of whether these cereals are completely safe and gluten-free for allergy sufferers remains open. One of the key steps in protein research is their efficient extraction. In this work, the Osborne sequential extraction method was developed and optimized using the response surface methodology in order to fractionate oat proteins. An optimized desirability of 0.986 was achieved with an extraction time of 4.7 min, a speed of 6, and a sample/solvent ratio of 5. The corresponding optimized responses were 8.7, 4.0, and 5.1% for the extraction yields of the avenin, avenalin, and albumin/globulin fractions, respectively. Further characterization of the extracts was carried out on 24 homogeneous and commercial oat samples via LC-MS/MS, targeting six potentially allergenic proteins. The avenin-E protein featured prominently, with relative contents of 60.7, 32.2, 58.0, and 59.8% in the total extract, avenin, avenalin, and albumin/globulin fractions, respectively, while the Avenin-3, ATI-2, avenin, SSG2, and SSG1 proteins in the total extract showed levels of 16.4, 9.3, 6.6, 4.8, and 2.2%, respectively. The preliminary results of an ELISA performed on the different fractions revealed low levels of gluten (from 1.24 ± 0.14 to 3.61 ± 0.16 mg/kg), which were well below the threshold limit of 20 mg/kg. These results support the hypothesis that oats can be a safe food for people suffering from cereal-related allergies. These results open the door to further studies into the comprehensive characterization of oat proteins.