SUMMARY: Protein language models (PLMs) have recently demonstrated potential to supplant classical protein database search methods based on sequence alignment, but are slower than common alignment-based tools and appear to be prone to a high rate of false labeling. Here, we present Neural Embeddings for Amino acid Relationships (NEAR), a method based on neural representation learning that is designed to improve both speed and accuracy of search for likely homologs in a large protein sequence database. NEAR's ResNet embedding model is trained using contrastive learning guided by trusted sequence alignments. It computes per-residue embeddings for target and query protein sequences, and identifies alignment candidates with a pipeline consisting of residue-level k-NN search and a simple neighbor aggregation scheme. Tests on a benchmark consisting of trusted remote homologs and randomly shuffled decoy sequences reveal that NEAR substantially improves accuracy relative to state-of-the-art PLMs, with lower memory requirements and faster embedding and search speed. While these results suggest that the NEAR model may be useful for standalone homology detection with increased sensitivity over standard alignment-based methods, in this manuscript, we focus on a more straightforward analysis of the model's value as a high-speed pre-filter for sensitive annotation. In that context, NEAR is at least 5x faster than the pre-filter currently used in the widely used profile hidden Markov model (pHMM) search tool HMMER3, and also outperforms the pre-filter used in our fast pHMM tool, nail. AVAILABILITY AND IMPLEMENTATION: NEAR is under an open-source license. Code and data curation instructions can be found at https://github.com/TravisWheelerLab/NEAR.

We introduce Artifact-Based Rendering (ABR), a framework of tools, algorithms, and processes that makes it possible to produce real, data-driven 3D scientific visualizations with a visual language derived entirely from colors, lines, textures, and forms created using traditional physical media or found in nature. A theory and process for ABR is presented to address three current needs: (i) designing better visualizations by making it possible for non-programmers to rapidly design and critique many alternative data-to-visual mappings; (ii) expanding the visual vocabulary used in scientific visualizations to depict increasingly complex multivariate data; (iii) bringing a more engaging, natural, and human-relatable handcrafted aesthetic to data visualization. New tools and algorithms to support ABR include front-end applets for constructing artifact-based colormaps, optimizing 3D scanned meshes for use in data visualization, and synthesizing textures from artifacts. These are complemented by an interactive rendering engine with custom algorithms and interfaces that demonstrate multiple new visual styles for depicting point, line, surface, and volume data. A within-the-research-team design study provides early evidence of the shift in visualization design processes that ABR is believed to enable when compared to traditional scientific visualization systems. Qualitative user feedback on applications to climate science and brain imaging support the utility of ABR for scientific discovery and public communication.

Do improvements in system performance demonstrated by batch evaluations confer the same benefit for real users? We carried out experiments designed to investigate this question. After identifying a weighting scheme that gave maximum improvement over the baseline in a non-interactive evaluation, we used it with real users searching on an instance recall task. Our results showed the weighting scheme giving beneficial results in batch studies did not do so with real users. Further analysis did identify other factors predictive of instance recall, including number of documents saved by the user, document recall, and number of documents seen by the user.