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.

Infections are a serious health concern worldwide, particularly in vulnerable populations such as the immunocompromised, elderly, and young. Advances in metagenomic sequencing availability, speed, and decreased cost offer the opportunity to supplement or even replace culture-based identification of pathogens with DNA sequence-based diagnostics. Adopting metagenomic analysis for clinical use requires that all aspects of the workflow are optimized and tested, including data analysis and computational time and resources. We tested the accuracy, sensitivity, and resource requirements of three top metagenomic taxonomic classifiers that use fast k-mer based algorithms: Centrifuge, CLARK, and KrakenUniq. Binary mixtures of bacteria showed all three reliably identified organisms down to 1% relative abundance, while only the relative abundance estimates of Centrifuge and CLARK were accurate. All three classifiers identified the organisms present in their default databases from a mock bacterial community of 20 organisms, but only Centrifuge had no false positives. In addition, Centrifuge required far less computational resources and time for analysis. Centrifuge analysis of metagenomes obtained from samples of VAP, infected DFUs, and FN showed Centrifuge identified pathogenic bacteria and one virus that were corroborated by culture or a clinical PCR assay. Importantly, in both diabetic foot ulcer patients, metagenomic sequencing identified pathogens 4-6 weeks before culture. Finally, we show that Centrifuge results were minimally affected by elimination of time-consuming read quality control and host screening steps.

UNLABELLED: CMap is a web-based tool for displaying and comparing maps of any type and from any species. A user can compare an unlimited number of maps, view pair-wise comparisons of known correspondences, and search for maps or for features by name, species, type and accession. CMap is freely available, can run on a variety of database engines and uses only free and open software components. AVAILABILITY: http://www.gmod.org/cmap