Large Molecular Datasets Research Articles

Friedmannodendron rodriguesiana, a new segregate of Dombeya Cav. (Dombeyoideae, Malvaceae), endemic to Rodrigues

ABSTRACT Dombeya rodriguesiana F.Friedmann is an endemic species to Rodrigues (Mascarene Archipelago, Indian Ocean). This species faces severe threats and is currently represented by at least two living individuals. We sequenced a second sample of this species to confirm that its phylogenetic placement renders Dombeya Cav. paraphyletic and, thus, its placement in Dombeya needs revision. The second sample was extracted from a cultivated clone of the last wild individual of D. rodriguesiana. Sequences from this individual for six markers (one nuclear, five plastid) were analyzed using the largest molecular dataset available on Dombeyoideae and Bayesian inference and maximum likelihood methods. Study of all known herbarium material of D. rodriguesiana, and other relevant material, allowed for a new morphological evaluation and description. Our results demonstrate that D. rodriguesiana is closely related to the Malagasy endemic genus, Andringitra Skema. We found that the unique combination of characters of a persistent epicalyx with a persistent perianth provides the synapomorphy for these two sister taxa, independent of the significant morphological differences between them (e.g. flower size, number, arrangement, sex expression). Furthermore, we describe useful field characters to differentiate D. rodriguesiana from the superficially similar D. ferruginea Cav. To maintain a meaningful classification that reflects monophyletic groups in the Dombeyoideae, we propose the establishment of a new genus and combination to accommodate this rare species.

Fossil-calibrated molecular clock data enable reconstruction of steps leading to differentiated multicellularity and anisogamy in the Volvocine algae.

Throughout its nearly four-billion-year history, life has undergone evolutionary transitions in which simpler subunits have become integrated to form a more complex whole. Many of these transitions opened the door to innovations that resulted in increased biodiversity and/or organismal efficiency. The evolution of multicellularity from unicellular forms represents one such transition, one that paved the way for cellular differentiation, including differentiation of male and female gametes. A useful model for studying the evolution of multicellularity and cellular differentiation is the volvocine algae, a clade of freshwater green algae whose members range from unicellular to colonial, from undifferentiated to completely differentiated, and whose gamete types can be isogamous, anisogamous, or oogamous. To better understand how multicellularity, differentiation, and gametes evolved in this group, we used comparative genomics and fossil data to establish a geologically calibrated roadmap of when these innovations occurred. Our ancestral-state reconstructions, show that multicellularity arose independently twice in the volvocine algae. Our chronograms indicate multicellularity evolved during the Carboniferous-Triassic periods in Goniaceae + Volvocaceae, and possibly as early as the Cretaceous in Tetrabaenaceae. Using divergence time estimates we inferred when, and in what order, specific developmental changes occurred that led to differentiated multicellularity and oogamy. We find that in the volvocine algae the temporal sequence of developmental changes leading to differentiated multicellularity is much as proposed by David Kirk, and that multicellularity is correlated with the acquisition of anisogamy and oogamy. Lastly, morphological, molecular, and divergence time data suggest the possibility of cryptic species in Tetrabaenaceae. Large molecular datasets and robust phylogenetic methods are bringing the evolutionary history of the volvocine algae more sharply into focus. Mounting evidence suggests that extant species in this group are the result of two independent origins of multicellularity and multiple independent origins of cell differentiation. Also, the origin of the Tetrabaenaceae-Goniaceae-Volvocaceae clade may be much older than previously thought. Finally, the possibility of cryptic species in the Tetrabaenaceae provides an exciting opportunity to study the recent divergence of lineages adapted to live in very different thermal environments.

AI will not give us precision medicine.

The completion of human DNA sequencing in the early 2000s initially generated widespread excitement and hope that it would revolutionize medicine. Over time, however, it revealed major limitations due to a lack of understanding of the highly complex genotype-phenotype pathway. Precision medicine has emerged as a response to these biotechnological innovations, tailoring treatments based on an array of new molecular and clinical "omics" data. However, the large volume and heterogeneity of data available today requires the use of dedicated and highly efficient computational analyses. Widely used today are artificial intelligence techniques (such as machine learning) based on artificial neural networks, i.e., a mathematical model of how biological neurons work. Here, we show that artificial neural networks have nothing to do with biology, although their popularity is largely due to their alleged ability to simulate the human brain. Furthermore, we argue that the analysis of large molecular datasets cannot be left to the computational side alone, i.e., to be exclusively data-driven, but on the contrary must meet the challenge of integrating data and expertise, of getting clinicians and data analysts to work together to take into account the absolute and ineradicable uniqueness of each patient's characteristics.

Creation of Polymer Datasets with Targeted Backbones for Screening of High-Performance Membranes for Gas Separation.

A simple approach was developed to computationally construct a polymer dataset by combining simplified molecular-input line-entry system (SMILES) strings of a targeted polymer backbone and a variety of molecular fragments. This method was used to create 14 polymer datasets by combining seven polymer backbones and molecules from two large molecular datasets (MOSES and QM9). Polymer backbones that were studied include four polydimethylsiloxane (PDMS) based backbones, poly(ethylene oxide) (PEO), poly(allyl glycidyl ether) (PAGE), and polyphosphazene (PPZ). The generated polymer datasets can be used for various cheminformatics tasks, including high-throughput screening for gas permeability and selectivity. This study utilized machine learning (ML) models to screen the polymers for CO2/CH4 and CO2/N2 gas separation using membranes. Several polymers of interest were identified. The results highlight that employing an ML model fitted to polymer selectivities leads to higher accuracy in predicting polymer selectivity compared to using the ratio of predicted permeabilities.

COMPREHENSIVE MULTIPLATFORM ANALYSIS OF CDKN2A ALTERATIONS IN MENINGIOMAS

Abstract BTFC travel award recipient In meningiomas, CDKN2A/B deletions are associated with poor clinical outcomes but are exceeding rare in most cohorts (1-5% of cases). Large molecular datasets are therefore required to explore these deletions and their relationship to other CDKN2A alterations that may be more common, but also prognostic on a transcriptomic, epigenomic, and/or copy number level. METHODS: We used multidimensional molecular data of 560 meningioma samples from 5 independent cohorts to comprehensively interrogate the spectrum of CDKN2A alterations through DNA methylation, copy number variation, transcriptomics, and proteomics using an integrated molecular approach, and utilized preclinical models to validate our findings. RESULTS: Meningiomas with either CDKN2A/B deletions (partial or homozygous loss) or an intact CDKN2A gene locus but elevated mRNA expression (CDKN2Ahigh) both had poor clinical outcomes. Increased CDKN2A mRNA expression was a poor prognostic factor independent of deletion status. CDKN2A expression and p16 protein also progressively increased with tumor grade and more aggressive molecular and methylation groups. CDKN2Ahigh meningiomas and meningiomas with CDKN2A deletions were enriched for similar cell cycling pathways but dysregulated at different checkpoints (G1/S vs G2/M). p16 immunohistochemistry was unreliable in differentiating between meningiomas with and without CDKN2A deletions, but increased positivity was associated with mRNA expression. CDKN2Ahigh meningiomas were associated with gene hypermethylation, Rb-deficiency, and lack of response to CDK inhibition. CONCLUSIONS: In meningiomas, CDKN2A mRNA expression consistently increases with biological aggressiveness, is prognostic independent of copy number loss, and may be used as an important prognostic biomarker with therapeutic implications for resistance to CDK4/6 inhibitors clinically.

F.3 Comprehensive multiplatform analysis of CDKN2A alterations in meningiomas

Background: In meningiomas, CDKN2A/B deletions are associated with poor outcomes but are rare in most cohorts (1-5%). Large molecular datasets are therefore required to explore these deletions and their relationship to other prognostic CDKN2A alterations. Methods: We utilized multidimensional molecular data of 560 meningiomas from 5 independent cohorts to comprehensively interrogate the spectrum of CDKN2A alterations through DNA methylation, copy number variation, transcriptomics, and proteomics using an integrated molecular approach. Results: Meningiomas with either CDKN2A/B deletions (partial or homozygous loss) or an intact CDKN2A gene locus but elevated mRNA expression (CDKN2Ahigh) both had poor clinical outcomes. Increased CDKN2A mRNA expression was a poor prognostic factor independent of CDKN2A deletion. CDKN2A expression and p16 protein increased with tumor grade and more aggressive molecular and methylation groups. CDKN2Ahigh meningiomas and meningiomas with CDKN2A deletions were enriched for similar cell cycling pathways dysregulated at different checkpoints. p16 immunohistochemistry was unreliable in differentiating between meningiomas with and without CDKN2A deletions, but increased positivity was associated with increased mRNA expression. CDKN2Ahigh meningiomas were associated with gene hypermethylation, Rb-deficiency, and lack of response to CDK inhibition. Conclusions: These findings support the role of CDKN2A mRNA expression as a biomarker of clinically aggressive meningiomas with potential therapeutic implications.

Geometric Deep Learning for Molecular Crystal Structure Prediction.

We develop and test new machine learning strategies for accelerating molecular crystal structure ranking and crystal property prediction using tools from geometric deep learning on molecular graphs. Leveraging developments in graph-based learning and the availability of large molecular crystal data sets, we train models for density prediction and stability ranking which are accurate, fast to evaluate, and applicable to molecules of widely varying size and composition. Our density prediction model, MolXtalNet-D, achieves state-of-the-art performance, with lower than 2% mean absolute error on a large and diverse test data set. Our crystal ranking tool, MolXtalNet-S, correctly discriminates experimental samples from synthetically generated fakes and is further validated through analysis of the submissions to the Cambridge Structural Database Blind Tests 5 and 6. Our new tools are computationally cheap and flexible enough to be deployed within an existing crystal structure prediction pipeline both to reduce the search space and score/filter crystal structure candidates.

Abstract PD5-02: PD5-02 An Organoid Model System to Study Resistance Mechanisms, Predictive Biomarkers, and New Strategies to Overcome Therapeutic Resistance in Early-Stage Triple-Negative Breast Cancer

Abstract Background: While new treatments and improved subtyping schemas are anticipated to improve treatment response in triple-negative breast cancer (TNBC) patients, therapeutic resistance remains a significant challenge. Moreover, there is an urgent need for additional research model systems to study resistance and residual disease in breast cancer, including aggressive subtypes of breast cancer. Organoid culture is a promising technology used for growing breast cancer cells with high efficiency; however, the extent to which treatment resistance can be modeled using this system is unknown. This research used patient-derived organoid cultures in the context of computational analyses of large molecular and clinical datasets to study resistance mechanisms, biomarkers, and alternative treatment strategies to overcome drug resistance in early-stage TNBC. Methods: Organoid cultures were derived from breast tumor samples (taken from lumpectomy, mastectomy, or core biopsy samples), digested to the organoid level using collagenase, and grown in three dimensional cultures using a basement membrane extract and a fully-defined organoid medium (Dekkers et al. Nat Protoc 2021). An evaluation of all available I-SPY2 biomarker data (Wolf et al. Cancer Cell 2022) was performed focusing on genes, proteins, and pathways associated with resistance. These were then used to study whether resistance biomarkers identified in patient tumors are also present in organoids propagated from breast cancer post-treatment residual disease. To this end, bulk RNA sequencing data of organoids were normalized and merged with the TCGA dataset (Hoadley et al. Cell 2018) to enable analysis in a larger context, and immunofluorescence staining of organoids was performed. A high-throughput 386 anti-cancer drug compound screen and subsequent synergy testing with the most promising compounds were performed to identify and predict alternative treatment strategies. Additional assays to explore kinome activity in this organoid model are in progress. Results: A TNBC organoid biobank was established (n=31), which was enriched for inflammatory breast cancer (IBC; n=15), an aggressive form of breast cancer. Most organoids were derived from residual disease after neoadjuvant therapy. Bulk RNA sequencing analysis performed on 10 TNBC organoids revealed 3 subsets that were characterized predominantly by either normal-like/luminal androgen receptor or basal-like features or expressed distinct gene expression profiles, with IBC cases present in all 3 subsets. Intriguingly, the IBC organoids were characterized by higher expression of a number of immune-related signatures, despite an absence of clear immune cells in culture. A residual disease IBC/TNBC organoid resistant to chemotherapy was used to perform the 386-drug compound screen. The organoid model showed resistance to veliparib-cisplatin, consistent with the expression of gene/protein biomarkers predictive of drug resistance found in I-SPY2 (low PARPi7 levels and high pFOXO1 and pMEK1/2 expression). In addition, the screen identified multiple classes of inhibitors as promising synergistic/additive candidates for overcoming resistance to cisplatin. Conclusion: In this proof-of-principle study, we demonstrated the utility of matching I-SPY2 resistance biomarkers and signatures to residual disease tumor organoid cultures. We show that tumor organoid cultures modeling drug resistance states are a useful complement to existing research models of breast cancer and can be used for compound testing. We have developed a pipeline to propagate residual tumors from patients enrolled in I-SPY2 into organoid cultures to create a broader platform for preclinical drug testing informed by tumor biology with the ultimate goal of enabling faster, more successful translational studies and increased treatment options for resistant disease. Citation Format: Tam Binh V. Bui, Denise M. Wolf, Kaitlin Moore, Isaac S. Harris, Pravin Phadatare, Christina Yau, Lamorna A. Brown Swigart, Laura J. Esserman, Jean-Philippe Coppe, Julia Wulfkuhle, Emanuel F. Petricoin, Michael Campbell, Laura M. Selfors, Deborah A. Dillon, Beth Overmoyer, Filipa Lynce, Laura Van ’t Veer, Jennifer Rosenbluth. PD5-02 An Organoid Model System to Study Resistance Mechanisms, Predictive Biomarkers, and New Strategies to Overcome Therapeutic Resistance in Early-Stage Triple-Negative Breast Cancer [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr PD5-02.

Phylogenetic relationships of the North American catfishes (Ictaluridae, Siluriformes): Investigating the origins and parallel evolution of the troglobitic species

Insular habitats have played an important role in developing evolutionary theory, including natural selection and island biogeography. Caves are insular habitats that place extreme selective pressures on organisms due to the absence of light and food scarcity. Therefore, cave organisms present an excellent opportunity for studying colonization and speciation in response to the unique abiotic conditions that require extreme adaptations. One vertebrate family, the North American catfishes (Ictaluridae), includes four troglobitic species that inhabit the karst region bordering the western Gulf of Mexico. The phylogenetic relationships of these species have been contentious, and conflicting hypotheses have been proposed to explain their origins. The purpose of our study was to construct a time-calibrated phylogeny of Ictaluridae using first-occurrence fossil data and the largest molecular dataset on the group to date. We test the hypothesis that troglobitic ictalurids have evolved in parallel, thus resulting from repeated cave colonization events. We found that Prietella lundbergi is sister to surface-dwelling Ictalurus and that Prietella phreatophila + Trogloglanis pattersoni are sister to surface-dwelling Ameiurus, suggesting that ictalurids colonized subterranean habitats at least twice in evolutionary history. The sister relationship between Prietella phreatophila and Trogloglanis pattersoni may indicate that these two species diverged from a common ancestor following a subterranean dispersal event between Texas and Coahuila aquifers. We recovered Prietella as a polyphyletic genus and recommend P. lundbergi be removed from this genus. With respect to Ameiurus, we found evidence for a potentially undescribed species sister to A. platycephalus, which warrants further investigation of Atlantic and Gulf slope Ameiurus species. In Ictalurus, we identified shallow divergence between I. dugesii and I. ochoterenai, I. australis and I. mexicanus, and I. furcatus and I. meridionalis, indicating a need to reexamine the validity of each species. Lastly, we propose minor revisions to the intrageneric classification of Noturus including the restriction of subgenus Schilbeodes to N. gyrinus (type species), N. lachneri, N. leptacanthus, and N. nocturnus.

Neither slugs nor snails: a molecular reappraisal of the gastropod family Velutinidae

AbstractThe systematics of the marine mollusc family Velutinidae has long been neglected by taxonomists, mainly because their often internal and fragile shells offer no morphological characters. Velutinids are usually undersampled owing to their cryptic mantle coloration on the solitary, social or colonial ascidians on which they feed and lay eggs. In this study, we address the worldwide diversity and phylogeny of Velutinidae based on the largest molecular dataset (313 specimens) to date, accounting for > 50% of the currently accepted genera, coupled with morphological and ecological data. Velutinids emerge as a diverse group, encompassing four independent subfamily-level lineages, two of which are newly described herein: Marseniopsinae subfam. nov. and Hainotinae subfam. nov. High diversity was found at genus and species levels, with two newly described genera (Variolipallium gen. nov. and Pacifica gen. nov.) and ≥ 86 species in the assayed dataset, 58 of which are new to science (67%). Velutinidae show a remarkable morphological plasticity in shell morphology, mantle extension and chromatic patterns. This variability is likely to be the result of different selective forces, including habitat, depth and trophic interactions.

Improving VAE based molecular representations for compound property prediction

Collecting labeled data for many important tasks in chemoinformatics is time consuming and requires expensive experiments. In recent years, machine learning has been used to learn rich representations of molecules using large scale unlabeled molecular datasets and transfer the knowledge to solve the more challenging tasks with limited datasets. Variational autoencoders are one of the tools that have been proposed to perform the transfer for both chemical property prediction and molecular generation tasks. In this work we propose a simple method to improve chemical property prediction performance of machine learning models by incorporating additional information on correlated molecular descriptors in the representations learned by variational autoencoders. We verify the method on three property prediction tasks. We explore the impact of the number of incorporated descriptors, correlation between the descriptors and the target properties, sizes of the datasets etc. Finally, we show the relation between the performance of property prediction models and the distance between property prediction dataset and the larger unlabeled dataset in the representation space.

Quantitatively defining species boundaries with more efficiency and more biological realism

We introduce a widely applicable species delimitation method based on the multispecies coalescent model that is more efficient and more biologically realistic than existing methods. We extend a threshold-based method to allow the ancestral speciation rate to vary through time as a smooth piecewise function. Furthermore, we introduce the cutting-edge proposal kernels of StarBeast3 to this model, thus enabling rapid species delimitation on large molecular datasets and allowing the use of relaxed molecular clock models. We validate these methods with genomic sequence data and SNP data, and show they are more efficient than existing methods at achieving parameter convergence during Bayesian MCMC. Lastly, we apply these methods to two datasets (Hemidactylus and Galagidae) and find inconsistencies with the published literature. Our methods are powerful for rapid quantitative testing of species boundaries in large multilocus datasets and are implemented as an open source BEAST 2 package called SPEEDEMON.

Kenneth Wade. 13 October 1932—16 March 2014

Much of Ken Wade's scientific career was devoted to developing and exploiting the chemistries of boron, aluminium, gallium and indium (the Group 13 elements). He made significant contributions to both experimental and theoretical aspects of the subject and was a highly regarded teacher and attentive research supervisor. Ken was a modest and self-critical man who had a rare ability to recognize patterns in large chemical and molecular structure data sets and express the resulting generalizations in the form of simple rules that experimental chemists could appreciate and use to predict new compounds. Ken was fascinated by the structures of cluster compounds that are based on regular polyhedral shapes. Cluster molecules generally contain groups of metal atoms linked by metal–metal bonds and located on one or more spherical shells. Since the 1960s they have attracted the attention of both academic and industrial chemists. Understanding and predicting their three-dimensional shapes raised a significant intellectual challenge for chemists. The clusters also attracted attention because they occupy an intermediate position between co-ordination compounds based on a single metal atom and the parent metal, which has an infinite number of linked metal atoms. This provided chemists with a unique opportunity to explore the evolution of their electronic and chemical properties as the cluster size increases. Ken was the first to recognize the electronic relationships connecting closo- , nido- , arachno- and hypho - clusters. These clusters are inter-related by the successive removal of vertices from a regular closo- deltahedron, i.e. a polyhedron with all the faces triangular. His ability to connect the structures of these molecules to the number of valence electrons involved in skeletal bonding represented a significant contribution to chemical bonding theories. Ken communicated his conclusions in very clearly written papers and books. This made the complexities of cluster chemistry amenable to many and especially those who did not have a strong background in quantum chemistry. The skeletal electron counting rules, or Wade's Rules, which Ken pioneered, are now very much part of the fabric of modern cluster chemistry and form an important component of modern undergraduate inorganic chemistry courses. Wade's Rules may be applied to main group and transition metal carbonyl clusters as well as clusters containing both main group and transition metal fragments. Besides providing an accessible entry into cluster chemistry for students, the rules have been used creatively by many leading synthetic chemists to discover new classes of compounds.

Exploring Deep Learning of Quantum Chemical Properties for Absorption, Distribution, Metabolism, and Excretion Predictions.

Quantum mechanical (QM) descriptors of small molecules have wide applicability in understanding organic reactivity and molecular properties, but the substantial compute cost required for ab initio QM calculations limits their broad usage. Here, we investigate the use of deep learning for predicting QM descriptors, with the goal of enabling usage of near-QM accuracy electronic properties on large molecular data sets such as those seen in drug discovery. Several deep learning approaches have previously been benchmarked on a published data set called QM9, where 12 ground-state properties have been calculated for molecules with up to nine heavy atoms, limited to C, H, N, O, and F elements. To advance the work beyond the QM9 chemical space and enable application to molecules encountered in drug discovery, we extend the QM9 data set by creating a QM9-extended data set covering an additional ∼20,000 molecules containing S and Cl atoms. Using this extended set, we generate new deep learning models as well as leverage ANI-2x models to provide predictions on larger, more diverse molecules common in drug discovery, and we find the models estimate 11 of 12 ground-state properties reasonably. We use the predicted QM descriptors to augment graph convolutional neural network (GCNN) models for selected ADME end points (rat microsomal clearance, hepatic clearance, total clearance, and P-glycoprotein efflux) and found varying degrees of performance improvement compared to nonaugmented GCNN models, including pronounced improvement in P-glycoprotein efflux prediction.

Benchmark Study of Alkane Molecular Chains

Alkanes serve an important role as benchmark system in molecular electronics. However, a large variation in the conductance values is reported in the literature. To better understand these fluctuations, in this study we measure large molecular data sets (up to 100 000 breaking traces) of a series of alkanes with different lengths and anchoring groups using the mechanically controlled break junction (MCBJ) technique. Employing an unsupervised learning algorithm, we investigate both the time evolution and the distance dependence of the measured traces. For all the molecules considered, we have been able to identify the single-molecule conductance value for the fully stretched molecular configuration. For alkanedithiols, the corresponding extracted β decay constant of 1.05 ± 0.08 per CH2 group agrees well with literature values. In the case of the stronger thiol bonding, additional peaks in the conductance histograms are found, suggesting the formation of molecular junctions containing a single molecule plus additional gold/molecule unit(s). The results shine light on the dispersion in reported conductance values and show that the evolution of the molecule as a function of stretching and time contains crucial information in determining the molecular junction configuration in MCBJs.

Assessment of the Efficiency of Chemical and Thermochemical Depolymerization Methods for Lignin Valorization: Principal Component Analysis (PCA) Approach.

Energy demand and the use of commodity consumer products, such as chemicals, plastics, and transportation fuels, are growing nowadays. These products, which are mainly derived from fossil resources and contribute to environmental pollution and CO2 emissions, will be used up eventually. Therefore, a renewable inexhaustible energy source is required. Plant biomass resources can be used as a suitable alternative source due to their green, clean attributes and low carbon emissions. Lignin is a class of complex aromatic polymers. It is highly abundant and a major constituent in the structural cell walls of all higher vascular land plants. Lignin can be used as an alternative source for fine chemicals and raw material for biofuel production. There are many chemical processes that can be potentially utilized to increase the degradation rate of lignin into biofuels or value-added chemicals. In this study, two lignin degradation methods, CuO–NaOH oxidation and tetramethyl ammonium hydroxide (TMAH) thermochemolysis, will be addressed. Both methods showed a high capacity to produce a large molecular dataset, resulting in tedious and time-consuming data analysis. To overcome this issue, an unsupervised machine learning technique called principal component analysis (PCA) is implemented.

Abstract 867: Incidence of molecular alterations in KRAS and other known cancer genes in patients with pancreatic cancer assessed with a commercial genomic profiling panel compared to TCGA results

Abstract Background: The Cancer Genome Atlas (TCGA) characterizes somatic mutations of pancreatic cancer (PC) patients (pts) using whole-exome sequencing (WES, mean coverage 405x) of fresh frozen tumor tissues and matched blood samples. In addition, it includes a confirmation of the key genomic findings by a targeted sequencing panel (~644x) and microfluidic PCR-based sequencing (~30,000x). For the vast majority of patients who do not undergo resection, however genomic characterization of formalin-fixed, paraffin-embedded (FFPE) biopsy samples is the only available assessment. In order to assess this NGS testing process at Baylor College of Medicine, we sought to confirm that routinely collected diagnostic FFPE specimens could be used for sequencing without compromising the detection rates of key gene alterations in PC pts. Methods: Matched tumor and normal specimens (peripheral blood or saliva) from 59 Baylor College of Medicine PC pts were sequenced using Tempus|xT (Tempus Labs, Chicago, IL), a 648-gene DNA sequencing panel (avg. depth of 500x) integrated with whole-transcriptome RNA sequencing. Retrospective analysis of detection rates of all clinically reported pathogenic mutations was performed using LENS, a proprietary application that allows interrogation of large de-identified molecular and clinical datasets. Results: The tumor cellularity of the 59 FFPE samples as assessed by board-certified pathologists ranged from 10% to 80%. Approximately 88% (52/59) of these samples had detectable KRAS alterations, relative to 93% identified in the TCGA database. KRAS G12D, G12V, and G12R were the most frequent findings (29/59, 11/59, and 8/59, respectively). The average KRAS variant allele fraction (VAF) was 16.87% (range 1.7%- 42.2%). Alterations in CDKN2A were detected with a noticeably higher incidence as compared to the TCGA database, 37.2% (22/59) vs 17.4%, respectively. Conversely, TP53 and SMAD4 had similar detection incidence between the two platforms. In addition, 13.5% (8/59) of PC pts had a somatic or germline mutation in the homologous recombination (HR) pathway, with 4 pts having a BRCA2 alteration (6.7%). Finally, one patient was tumor mutational burden (TMB)-High and microsatellite instability (MSI)-High, both of which are tumor-agnostic indications for checkpoint inhibition. Conclusions: An analysis of PC FFPE tissues using the Tempus|xT panel at a clinical depth of 500x revealed a detection rate of KRAS, TP53, and SMAD4 alterations comparable to the TCGA database. Citation Format: Alexander D. Le, Bing Zhang, Kim C. Worley, David Wheeler, Benjamin L. Musher, Heidi Dowst, Amy Wood, Jeff Schaeffer, Nick Olsen, Lorenzo Grego, Matt Sonnleitner, William E. Fisher. Incidence of molecular alterations in KRAS and other known cancer genes in patients with pancreatic cancer assessed with a commercial genomic profiling panel compared to TCGA results [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 867.

Total Evidence Phylogenetic Analysis Supports New Morphological Synapomorphies for Bovidae (Mammalia, Artiodactyla)

Advances in gene sequencing technology have made it easier to generate large molecular datasets with novel DNA sequences for phylogenetic analysis. Because morphological data are difficult to collect and not required for molecular phylogenetic analyses, they are often excluded in studies of the systematic relationships of extant taxa. This fact is especially apparent in the Bovidae, the highly diverse, widespread clade of hoofed mammals most often characterized by the presence of permanent bony horn cores covered with keratin sheaths. Analyses of molecular data have reconstructed well-supported phylogenetic relationships within the clade. However, morphological data are also required to integrate fossil taxa into these studies, and may support different topologies when they are included in total evidence phylogenies. In this study, I performed a maximum likelihood phylogenetic analysis of a total evidence dataset including mitochondrial genomes, nine nuclear genes, and 196 morphological characters. The dataset comprises 156 species, 13 of which are fossil taxa, one of the most complete analyses of the family to date, and the first time many of the fossil species have been included in a total evidence analysis. Character optimizations on this topology produced seven synapomorphic morphological characters for Bovidae and multiple characters for each tribe. These analyses support the use of total evidence phylogenetics as a means of uncovering morphological characters that may serve as new synapomorphies and elucidate the systematic relationships of fossil species.

Introducing block design in graph neural networks for molecular properties prediction

The number of states required for describing a many-body quantum system increases exponentially with the number of particles; thus, it is time- and effort-consuming to exactly calculate molecular properties. Herein, we propose a deep learning algorithm named block-based graph neural network (BGNN) as an approximate solution. The algorithm can be understood as a representation learning process to extract useful interactions between a target atom and its neighboring atomic groups. Compared to other graph model variants, BGNN achieved the smallest mean absolute errors in most tasks on two large molecular datasets, QM9 and Alchemy. Our advanced machine learning method exhibits general applicability and can be readily employed for bioactivity prediction and other tasks relevant to drug discovery and materials design.

<p><strong>Fleas are parasitic scorpionflies</strong></p>

Fleas (Siphonaptera) are medically important blood-feeding insects responsible for spreading pathogens such as plague, murine typhus, and myxomatosis. The peculiar morphology of fleas resulting from their specialised ectoparasitic lifestyle has meant that the phylogenetic position of this diverse and medically important group has remained one of the most persistent problems in insect evolution. Here we test competing hypotheses on the contentious evolutionary relationships of fleas and antliophoran insects using the largest molecular dataset available to date consisting of over 1,400 protein-coding genes, and a smaller mitogenome and Sanger sequence alignment of 16 genes. By removing ambiguously aligned sequence regions and using site-heterogeneous models, we consistently recover fleas nested within scorpionflies (Mecoptera) as sister to the relictual southern hemisphere family Nannochoristidae. Topology tests accounting for compositional heterogeneity strongly favour the proposed topology over previous hypotheses of antliophoran relationships. This clade is diagnosed by shared morphological characters of the head and sperm pump. Fleas may no longer be regarded as a separate insect order and we propose that Siphonaptera should be treated as an infraorder within Mecoptera, reducing the number of extant holometabolan insect orders to ten.