Search NASA⌕ Search

SEARCH · Search NASA

Results for “LABELLING”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,279 records · Page 71
All sourcesNASA NTRSDOE OSTI
DOE OSTI2018

pnnl/Haskell-IDF-Executable

Executables for working with EnergyPlus Input Files (IDF) This project aims to provide a suite of executables for working with EnergyPlus Input Files (IDF). The project is developed using the Haskell programming language (c.f., https://www.haskell.org/). The current release (version 0.1.0.0) includes a utility that "pretty prints" the input file (i.e., that decodes and then, immediately, re-encodes and displays the input file, viz., that sanitizes the input) and a utility that converts the input file into a labeled graph in the DOT language (c.f., https://www.graphviz.org/doc/info/lang.html) for subsequent rendering using Graphviz Visualization Software (c.f., https://www.graphviz.org/). An example workflow is to pipe the input file into the pretty printer, to pipe the result into the labeled graph converter, and then to pipe the result into Graphviz, yielding an image file that depicts the hierarchical and referential relationships between EnergyPlus Data Dictionary (IDD) classes that are detected within the input file.

Borkum, Mark↗
DOE OSTI2017

pnnl/safekit

Some of the most pressing machine learning applications such as cyber security and object recognition lack enough ground-truth training data to build a classifier. We solve this problem by using unsupervised deep learning techniques to determine when data were anomalous or deviated from the norm. Traditional methods require either specific context dependent expertise to construct models, or prior examples of events and large amounts of well-labeled data to perform machine learning. Some cutting-edge deep learning approaches have been used to characterize complex events, but these have not been applied to the streaming environment and require a great deal of data. We demonstrated the use of an autoencoder and relevant featurization techniques to both learn a feature space and then identify anomalous portions without the aid of labeled data or domain knowledge. In this way, we demonstrated the detection of anomalous features without domain knowledge or tagged examples, and delivered these features to users.

Central, PNNL Developer↗
DOE OSTI2025

Value of Information App (Value of Information App for Binary Geothermal Decisions and Binary Geothermal Possibilities) (Negative/Positive) [SWR-25-15]

Code base to run Streamlit Value of Information App for binary decision with geothermal techno economics. An open-source VOI app that models binary decisions (e.g. do something (drill) or walk away (do nothing)) and binary geothermal scenarios (positive or negative) has been developed. Users can input their anticipated economic values (profits or losses) directly into the value matrix to represent all four combinations of these actions and geothermal possibilities. VOI in general requires probabilities to be assigned for “probability of success”, or probability of experiencing a positive geothermal scenario versus negative. The users of the App can toggle this probability of success both in the demo problem and in the Value of Imperfect Information problem. The VOI App allows users to upload their own labeled data to evaluate how well it allows them to distinguish between positive versus negative sites. We have been using IGNENIOUS data to test and demonstrate; industry members have prepared their own labeled data, and have present their examples from diverse use cases at a conference workshop. The VOI App is open to the public at: https://voigeothermalrising.streamlit.app

Trainor-Guitton, Whitney [National Renewable Energ↗
DOE OSTI2026

matsim-agents v1.0

matsim-agents is a multi-agent AI framework for atomistic materials simulation and discovery. It orchestrates large language models (LLMs), machine-learned interatomic potentials (MLIPs), and DFT codes into a single agentic loop running on laptops and DOE leadership-class supercomputers. MULTI-AGENT ORCHESTRATION A LangGraph state machine with three nodes: a Planner that converts a natural-language research objective into structured tasks; an Executor that dispatches atomistic tools and loops until the queue is empty; and an Analyst that summarizes results into a human-readable report. State is checkpointed after every step and human-in-the-loop gates can be inserted at any edge. HYPOTHESIS-DRIVEN DISCOVERY CHAT An interactive REPL (matsim-agents chat) that couples LLM dialogue with atomistic simulation. Chemical formulas are automatically detected in conversation turns and trigger a full crystal-phase exploration: structure generation → relaxation → stability scoring → result injection back into the conversation, creating a closed hypothesis-refinement loop. CRYSTAL PHASE ENUMERATION Given a composition, the phase explorer enumerates prototypes by stoichiometry: elemental (fcc/bcc/hcp/sc/diamond), binary 1:1 (rocksalt/CsCl/zincblende/ wurtzite/fluorite/rutile), ternary 1:1:3 (cubic perovskite), ternary 1:2:4 (perovskite + spinel), quaternary 1:1:2:6 (Fm-3m double perovskite). 2-D prototypes (graphene, h-BN, MoS2 2H/1T) and multilayer stacking are also supported via --include-2d and --num-layers. SUPERCELL GENERATION AND SITE DECORATION Auto-tiling to a minimum atom count (--min-atoms), explicit NxNxN tiling (--supercell), symmetry-distinct site decorations (--n-orderings), and isotropic lattice-scale sweeps (--lattice-scales) for volume bracketing. MLFF RELAXATION AND STABILITY SCORING HydraGNN (multi-headed GNN) drives structure relaxation via ASE with FIRE, BFGS, or BFGSLineSearch. Stability output: delta-E/atom ranking across phases and a max-residual-force dynamical-stability proxy. Other MLIPs (MACE, NequIP, Orb) can be plugged in through the same interface. DFT BACKENDS Quantum ESPRESSO pw.x and VASP 6.6 are first-class labellers. Both have validated GPU builds and SLURM/PBS launchers for three DOE platforms: Frontier (AMD MI250X, ROCm), Aurora (Intel PVC, oneAPI), Perlmutter (NVIDIA A100, CUDA). QE produces ~100 binaries (pw.x, ph.x, epw.x, ...). VASP supports scf, relax, vc-relax, and vc-relax-shape run types. ACTIVE-LEARNING LOOP matsim-agents al run CONFIG.yaml drives an iterative HydraGNN-DFT loop: MD generates candidates → ensemble/MC-dropout uncertainty selects the most informative → DFT labels them in parallel inside one allocation → dataset grows → HydraGNN retrains → repeat. DFT backend is a single YAML toggle (dft.backend: vasp | qe). LLM-generated seed structures are supported (no curated POSCAR library needed). Config uses ${VAR}, ${VAR:-default}, ${VAR:?msg} shell-style substitution for cross-user/cross-site portability. LLM BACKENDS Ollama (local, default), vLLM (HPC multi-GPU serving), OpenAI, Anthropic, HuggingFace Transformers+Accelerate. Selected at runtime via flag or env var with no code changes. HPC PORTABILITY Same Python entry points run on Frontier (ROCm 7.2), Aurora (oneAPI), and Perlmutter (CUDA 12). DFT and ML stacks are never co-loaded in the same shell; they couple through the scheduler and filesystem. Advanced multi-node launchers (serve, discovery-chat, single-relaxation, active-learning, QE warm-start) are provided for all three platforms. CODABENCH COMPETITION BUNDLE A self-contained benchmark: 159 atomistic test structures across 11 material classes, 5 tasks (formation energy, forces, ML relaxation, AI-DFT relaxation, phase stability ranking), public/private leaderboard split (30/70), and four ready-to-run baselines: MACE-MP-0, HydraGNN, UMA, AllScAIP.

Lupo Pasini, Massimiliano [Oak Ridge National Labo↗
DOE OSTI2025

A knowledge-informed large language model framework for U.S. nuclear power plant shutdown initiating event classification for probabilistic risk assessment

Identifying and classifying shutdown initiating events (SDIEs) is critical for developing shutdown probabilistic risk assessment for nuclear power plants. Existing computational approaches cannot achieve satisfactory performance due to the challenges of unavailable large, labeled datasets, imbalanced event types, and label noise. To address these challenges, we propose a hybrid pipeline that integrates a knowledge-informed machine learning model to prescreen non-SDIEs and a large language model (LLM) to classify SDIEs into four types. In the prescreening stage, we proposed a set of 44 SDIE text patterns that consist of the most salient keywords and phrases from six SDIE types. Text vectorization based on the SDIE patterns generates feature vectors that are highly separable by using a simple binary classifier. The second stage builds Bidirectional Encoder Representations from Transformers (BERT)-based LLM, which learns generic English language representations from self-supervised pretraining on a large dataset and adapts to SDIE classification by fine-tuning it on an SDIE dataset. The proposed approaches are evaluated on a dataset with 10,928 events using precision, recall ratio, F 1 score, and average accuracy. In conclusion, the results demonstrate that the prescreening stage can exclude more than 97% non-SDIEs, and the LLM achieves an average accuracy of 95.1% for SDIE classification.

99 - GENERAL AND MISCELLANEOUS↗
DOE OSTI2018

Quantitative phosphoproteomic analysis reveals reciprocal activation of receptor tyrosine kinases between cancer epithelial cells and stromal fibroblasts

Background Cancer-associated fibroblasts (CAFs) are one of the most important components of tumor stroma and play a key role in modulating tumor growth. However, a mechanistic understanding of how CAFs communicate with tumor cells to promote their proliferation and invasion is far from complete. A major reason for this is that most current techniques and model systems do not capture the complexity of signal transduction that occurs between CAFs and tumor cells. Methods In this study, we employed a stable isotope labeling with amino acids in cell culture (SILAC) strategy to label invasive breast cancer cells, MDA-MB-231, and breast cancer patient-derived CAF cells. We used an antibody-based phosphotyrosine peptide enrichment method coupled to LC–MS/MS to catalog and quantify tyrosine phosphorylation-mediated signal transduction events induced by the bidirectional communication between patient-derived CAFs and tumor cells. Results We discovered that distinct signaling events were activated in CAFs and in tumor epithelial cells during the crosstalk between these two cell types. We identified reciprocal activation of a number of receptor tyrosine kinases including EGFR, FGFR1 and EPHA2 induced by this bidirectional communication. Conclusions Our study not only provides insights into the mechanisms of the interaction between CAFs and tumor cells, but the model system described here could be used as a prototype for analysis of intercellular communication in many different tumor microenvironments.

Biochemistry & Molecular Biology↗
DOE OSTI2020

Predicting potential adverse events using safety data from marketed drugs

Abstract Background While clinical trials are considered the gold standard for detecting adverse events, often these trials are not sufficiently powered to detect difficult to observe adverse events. We developed a preliminary approach to predict 135 adverse events using post-market safety data from marketed drugs. Adverse event information available from FDA product labels and scientific literature for drugs that have the same activity at one or more of the same targets, structural and target similarities, and the duration of post market experience were used as features for a classifier algorithm. The proposed method was studied using 54 drugs and a probabilistic approach of performance evaluation using bootstrapping with 10,000 iterations. Results Out of 135 adverse events, 53 had high probability of having high positive predictive value. Cross validation showed that 32% of the model-predicted safety label changes occurred within four to nine years of approval (median: six years). Conclusions This approach predicts 53 serious adverse events with high positive predictive values where well-characterized target-event relationships exist. Adverse events with well-defined target-event associations were better predicted compared to adverse events that may be idiosyncratic or related to secondary target effects that were poorly captured. Further enhancement of this model with additional features, such as target prediction and drug binding data, may increase accuracy.

Daluwatte, Chathuri↗
DOE OSTI2026

Single-cell proteomics of Arabidopsis leaf mesophyll reveals dynamic protein responses to water-deficit stress

Background The application of single-cell omics tools to biological systems can provide unique insights into diverse cellular populations and their heterogeneous responses to internal and external perturbations. Thus far, most single-cell studies in plant systems have been limited to RNA-sequencing approaches, which only provide indirect readouts of cellular functions. Results Here, we present a single-cell proteomics workflow for plant cells that integrates tape-sandwich protoplasting, piezoelectric cell sorting, nanoPOTS sample preparation, and ion mobility-based MS data acquisition method for label-free single-cell proteomics analysis of Arabidopsis leaf mesophyll cells. From a single leaf protoplast, over 3,000 proteins were quantified with high precision. The workflow is demonstrated to identify stress associated changes in protein abundance by analyzing 117 protoplasts from well-watered and water-deficit stressed plants. Additionally, we describe a new approach for constructing covarying protein networks at the single-cell level and demonstrate how single-cell protein covariation analysis can reveal previously unrecognized protein functions while also capturing stress-induced changes in protein–protein dynamics. Conclusions The label-free scProteomic approach presented here represents a significant advance through the demonstration of a facile protoplast isolation method combined with deep and precise proteomic coverage of Arabidopsis leaf mesophyll cell types. We believe this study will serve as an informative reference to future plant scProteomic investigations.

Arabidopsis↗
DOE OSTI2021

Solid-state NMR of unlabeled plant cell walls: high-resolution structural analysis without isotopic enrichment

Abstract Background Multidimensional solid-state nuclear magnetic resonance (ssNMR) spectroscopy has emerged as an indispensable technique for resolving polymer structure and intermolecular packing in primary and secondary plant cell walls. Isotope ( 13 C) enrichment provides feasible sensitivity for measuring 2D/3D correlation spectra, but this time-consuming procedure and its associated expenses have restricted the application of ssNMR in lignocellulose analysis. Results Here, we present a method that relies on the sensitivity-enhancing technique Dynamic Nuclear Polarization (DNP) to eliminate the need for 13 C-labeling. With a 26-fold sensitivity enhancement, a series of 2D 13 C– 13 C correlation spectra were successfully collected using the unlabeled stems of wild-type Oryza sativa (rice). The atomic resolution allows us to observe a large number of intramolecular cross peaks for fully revealing the polymorphic structure of cellulose and xylan. NMR relaxation and dipolar order parameters further suggest a sophisticated change of molecular motions in a ctl1 ctl2 double mutant: both cellulose and xylan have become more dynamic on the nanosecond and microsecond timescale, but the motional amplitudes are uniformly small for both polysaccharides. Conclusions By skipping isotopic labeling, the DNP strategy demonstrated here is universally extendable to all lignocellulose materials. This time-efficient method has landed the technical foundation for understanding polysaccharide structure and cell wall assembly in a large variety of plant tissues and species.

09 BIOMASS FUELS↗
DOE OSTI2025

Quantitative stable isotope probing (qSIP) and cross-domain networks reveal bacterial-fungal interactions in the hyphosphere

Interactions between fungi and bacteria have the potential to substantially influence soil carbon dynamics in soil, but we have yet to fully identify these interactions and partners in their natural environment. In this study, we stacked two powerful methods, 13 C quantitative stable isotope probing (qSIP) and cross-domain co-occurrence network, to identify interacting fungi and bacteria in a California grassland soil. We used in-field whole plant 13 CO 2 labeling along with sand-filled ingrowth bags (that trap fungi and hyphae-associated bacteria) to amplify the signal of fungal-bacterial interactions, separate from the bulk soil background. We found a total of 54 bacterial ASVs and 9 fungal OTUs that were significantly 13 C-enriched. These were saprotrophic and biotrophic fungi, and motile, sometimes predatory bacteria. Among these, 70% of all 13 C-enriched bacteria identified were motile. Notably, we detected fungal-bacterial network links between a fungal OTU of the genus Alternaria and several bacterial ASVs of the genera Bacteriovorax, Mucilaginibacter, and Flavobacterium, providing empirical evidence of their direct interactions through C exchange. We observed a strong positive co-occurrence pattern between predatory bacteria of the phylum Bdellovibrionota and fungal OTUs, suggesting the transfer of C across the soil food web. To date, our ability to associate microbial co-occurrence network patterns with biological interactions is limited, but the incorporation of qSIP allowed us to more precisely detect interacting partners by narrowing in on the taxa that were actively incorporating plant-fixed, fungal-transported labeled substrates. Together, these approaches can help build a mechanistic understanding of the complex nature of fungal-bacterial interactions in soil.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Quantum cluster algorithm for data classification

Abstract We present a quantum algorithm for data classification based on the nearest-neighbor learning algorithm. The classification algorithm is divided into two steps: Firstly, data in the same class is divided into smaller groups with sublabels assisting building boundaries between data with different labels. Secondly we construct a quantum circuit for classification that contains multi control gates. The algorithm is easy to implement and efficient in predicting the labels of test data. To illustrate the power and efficiency of this approach, we construct the phase transition diagram for the metal-insulator transition of VO 2 , using limited trained experimental data, where VO 2 is a typical strongly correlated electron materials, and the metallic-insulating phase transition has drawn much attention in condensed matter physics. Moreover, we demonstrate our algorithm on the classification of randomly generated data and the classification of entanglement for various Werner states, where the training sets can not be divided by a single curve, instead, more than one curves are required to separate them apart perfectly. Our preliminary result shows considerable potential for various classification problems, particularly for constructing different phases in materials.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

The Value of {sup 99m}Tc-PSMA SPECT/CT-Guided Surgery for Identifying and Locating Lymph Node Metastasis in Prostate Cancer Patients

Background: This study evaluated the effect of technetium-99m ({sup 99m}Tc)-labeled prostate-specific membrane antigen (PSMA)-based image-guided surgery on the oncologic outcomes for patients with primary or recurrent prostate cancer (PCa). Methods: This study retrospectively analyzed 54 consecutive patients with PCa who underwent {sup 99m}Tc-labeled PSMA-based image-guided surgery between January 2016 and September 2017. These patients received a radical prostatectomy (RP) with pelvic lymph node dissection (PLND) or salvage lymph node dissection (sLND). The resected specimens were compared with findings of postoperative histologic analysis. The responses to the treatment were recorded during the follow-up period. Results: In 31 patients, PSMA single-photon emission computerized tomography (SPECT) and computed tomography (CT) could find 52 suspicious lymph node metastases (LNMs). With the help of PSMA SPECT/CT, 12 patients with recurrence received sLND, 19 primary PCa patients received RP with extended PLND, and 23 primary PCa patients received RP with standard PLND. The findings showed that PSMA SPECT/CT could detect LNMs with high sensitivity and specificity. In six patients, PSMA SPECT/CT could find more LNMs that were not found by MRI and help to modify the extent of lymphadenectomy. At the latest follow-up evaluation, 39 patients showed a biochemical response (BR), 9 patients showed a biochemical recurrence (BCR) after BR, and 6 patients never exhibited BR. The patients who received RP with standard PLND or extended PLND had a better prostate-specific antigen (PSA) response than the patients who received sLND. The patients with pelvic LNMs also had a better PSA response than the patients with retroperitoneal LNMs. Conclusions: This study showed that {sup 99m}Tc-PSMA SPECT/CT-guided surgery can remove more LNMs than conventional imaging with high sensitivity and specificity and delay disease progression in PCa patients.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2021

CATMoS: Collaborative Acute Toxicity Modeling Suite

Background: Humans are exposed to tens of thousands of chemical substances that need to be assessed for their potential toxicity. Acute systemic toxicity testing serves as the basis for regulatory hazard classification, labeling, and risk management. However, it is cost- and time-prohibitive to evaluate all new and existing chemicals using traditional rodent acute toxicity tests. In silico models built using existing data facilitate rapid acute toxicity predictions without using animals. Objectives: The U.S. Interagency Coordinating Committee on the Validation of Alternative Methods Acute Toxicity Workgroup organized an international collaboration to develop in silico models for predicting acute oral toxicity based on five different endpoints: LD50 value, U.S. Environmental Protection Agency hazard categories, Globally Harmonized System for Classification and Labelling hazard categories, very toxic chemicals (LD50 =50 mg/kg), and non-toxic chemicals (LD50 >2000 mg/kg). Methods: An acute oral toxicity data inventory for 11,992 chemicals was compiled, split into training and evaluation sets, and made available to 35 participating international research groups that submitted a total of 139 predictive models. Predictions that fell within the applicability domains of the submitted models were evaluated using external validation sets. These were then combined into consensus models to leverage strengths of individual approaches. Results: The resulting consensus predictions, which leverage the collective strengths of each individual model, form the Collaborative Acute Toxicity Modeling Suite (CATMoS). CATMoS demonstrated high performance in terms of accuracy and robustness when compared to in vivo results. Discussion: CATMoS is being evaluated by regulatory agencies for its utility and applicability as a potential replacement for in vivo rat acute oral toxicity studies. CATMoS predictions for over 800,000 chemicals have been made available via the NTP’s Integrated Chemical Environment. The models are also implemented in a free, standalone open-source tool, OPERA, which allows predictions of new and untested chemicals to be made.

63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTAN↗
DOE OSTI2020

Data from Improved Annotation of Untargeted Metabolomics Data through Buffer Modifications That Shift Adduct Mass and Intensity

Annotation of untargeted high-resolution full-scan LC-MS metabolomics data remains challenging due to individual metabolites generating multiple LC-MS peaks arising from isotopes, adducts, and fragments. Adduct annotation is a particular challenge, as the same mass difference between peaks can arise from adduct formation, fragmentation, or different biological species. To address this, here we describe a buffer modification workflow (BMW) in which the same sample is run by LC-MS in both liquid chromatography solvent with 14NH3–acetate buffer and in solvent with the buffer modified with 15NH3–formate. Buffer switching results in characteristic mass and signal intensity changes for adduct peaks, facilitating their annotation. This relatively simple and convenient chromatography modification annotated yeast metabolomics data with similar effectiveness to growing the yeast in isotope-labeled media. Application to mouse liver data annotated both known metabolite and known adduct peaks with 95% accuracy. Overall, it identified 26% of ∼27 000 liver LC-MS features as putative metabolites, of which ∼2600 showed HMDB or KEGG database formula match. This workflow is well suited to biological samples that cannot be readily isotope labeled, including plants, mammalian tissues, and tumors.

Conversion↗
DOE OSTI2023

Layer-wise Imaging Dataset from Powder Bed Additive Manufacturing Processes for Machine Learning Applications (Peregrine v2021-03)

This dataset contains layer-wise powder bed images from three different powder bed printing technologies – laser powder bed fusion, electron beam powder bed fusion, and binder jetting. This dataset was collected and annotated using the internally-developed Peregrine software tool and is designed primarily to facilitate research into anomaly defect detection using image segmentation or similar techniques. A total of 20 layers are provided for each printing technology, with each layer of data consisting of one or more calibrated images and an annotation file containing pixel-wise ground truth labels. The ground truths were labeled by domain experts, typically printer technicians. Data in this release were collected at Oak Ridge National Laboratory between 2016 and 2020 and were compiled in March 2021.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Adaptive optical third-harmonic generation microscopy for in vivo imaging of tissues

Third-harmonic generation microscopy is a powerful label-free nonlinear imaging technique, providing essential information about structural characteristics of cells and tissues without requiring external labelling agents. In this work, we integrated a recently developed compact adaptive optics module into a third-harmonic generation microscope, to measure and correct for optical aberrations in complex tissues. Taking advantage of the high sensitivity of the third-harmonic generation process to material interfaces and thin membranes, along with the 1,300-nm excitation wavelength used here, our adaptive optical third-harmonic generation microscope enabled high-resolution in vivo imaging within highly scattering biological model systems. Examples include imaging of myelinated axons and vascular structures within the mouse spinal cord and deep cortical layers of the mouse brain, along with imaging of key anatomical features in the roots of the model plant Brachypodium distachyon. In all instances, aberration correction led to enhancements in image quality.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2018

Full Karyotype Interphase Cell Analysis

Aneuploidy seems to play not only a decisive role in embryonal development but also in tumorigenesis where chromosomal and genomic instability reflect a universal feature of malignant tumors. The cost of whole genome sequencing has fallen significantly, but it is still prohibitive for many institutions and clinical settings. No applied, cost-effective, and efficient technique has been introduced yet aiming at research to assess the ploidy status of all 24 different human chromosomes in interphases simultaneously, especially in single cells. Here, we present the selection of human probe DNA and a technique using multistep fluorescence in situ hybridization (FISH) employing four sets of six labeled FISH probes able to delineate all 24 human chromosomes in interphase cells. This full karyotype analysis approach will provide additional diagnostic potential for single cell analysis. The use of spectral imaging (SIm) has enabled the use of up to eight different fluorochrome labels simultaneously. Thus, scoring can be easily assessed by visual inspection, because SIm permits computer-assigned and distinguishable pseudo-colors to each probe during image processing. This enables full karyotype analysis by FISH of single-cell interphase nuclei.

Baumgartner, Adi↗
DOE OSTI2026

An improved dataset for predicting mammal infecting viruses from genetic sequence information

There have been several attempts to develop machine learning (ML) models to identify human infecting viruses from their genomic sequences, with varying degrees of success. Direct comparison between models is problematic, because these models are typically trained and evaluated on different datasets with alternative data splitting schemes, features, and model performance metrics. In this paper we present a standardized dataset of mammal infecting and non-infecting viral pathogens, refined from the previous work of Mollentze et al. to include the latest literature evidence, roughly doubling the number of curated host-virus records available to the community, and new host target labels, primate and mammal. The new host labels were included for several reasons, including previous reports that classification performance is better at broader taxonomic ranks and the idea that there may be more data for primate infection that might serve as a suitable proxy for zoonotic potential and avoidance of false positives for human infection due to absence of evidence. On this dataset, we report the performance of eight machine learning models for predicting mammal-infecting viruses from their genomic sequences. We find that randomly assigning cases in our improved dataset to training/testing sets, when compared to the original assignments into training/testing in Mollentze et al., increases the overall average ROC AUC of prediction of human infection from 0.663 ± 0.070 to 0.784 ± 0.013, consistent with the reduction in phylogenetic distance between train and test sets (relative entropy change from 3.00 to 0.08). The broadest host category of mammal infection can be predicted most reliably at 0.850 ± 0.020. We share our improved dataset and code to enable standardized comparisons of machine learning methods to predict human host infections. Overall, we have presented preliminary evidence that classification of virus host infection is more tractable at higher taxonomic ranks, that unsurprisingly reducing the phylogenetic distance between training and test sets can improve predictive performance, that peptide kmer features appear to be harmful to out of sample model performance, and we are left with the question of whether models for virus host prediction can reasonably be expected to perform well in out of sample scenarios given the likelihood that viruses do not share a common ancestor. Consistent with this concern, when the data is resampled such that there is no overlap between viral families in training and test sets (relative entropy > 24), models perform no better than random chance at prediction of human infection regardless of whether kmers are included (ROC AUC 0.50 ± 0.08) or not (ROC AUC 0.50 ± 0.04).

59 BASIC BIOLOGICAL SCIENCES↗