MetageNN: a memory-efficient neural network taxonomic classifier robust to sequencing errors and missing genomes.

BACKGROUND: With the rapid increase in throughput of long-read sequencing technologies, recent studies have explored their potential for taxonomic classification by using alignment-based approaches to reduce the impact of higher sequencing error rates. While alignment-based methods are generally slower, k-mer-based taxonomic classifiers can overcome this limitation, potentially at the expense of lower sensitivity for strains and species that are not in the database. RESULTS: We present MetageNN, a memory-efficient long-read taxonomic classifier that is robust to sequencing errors and missing genomes. MetageNN is a neural network model that uses short k-mer profiles of sequences to reduce the impact of distribution shifts on error-prone long reads. Benchmarking MetageNN against other machine learning approaches for taxonomic classification (GeNet) showed substantial improvements with long-read data (20% improvement in F1 score). By utilizing nanopore sequencing data, MetageNN exhibits improved sensitivity in situations where the reference database is incomplete. It surpasses the alignment-based MetaMaps and MEGAN-LR, as well as the k-mer-based Kraken2 tools, with improvements of 100%, 36%, and 23% respectively at the read-level analysis. Notably, at the community level, MetageNN consistently demonstrated higher sensitivities than the previously mentioned tools. Furthermore, MetageNN requires < 1/4th of the database storage used by Kraken2, MEGAN-LR and MMseqs2 and is > 7× faster than MetaMaps and GeNet and > 2× faster than MEGAN-LR and MMseqs2. CONCLUSION: This proof of concept work demonstrates the utility of machine-learning-based methods for taxonomic classification using long reads. MetageNN can be used on sequences not classified by conventional methods and offers an alternative approach for memory-efficient classifiers that can be optimized further.

TUGDA: task uncertainty guided domain adaptation for robust generalization of cancer drug response prediction from in vitro to in vivo settings.

MOTIVATION: Large-scale cancer omics studies have highlighted the diversity of patient molecular profiles and the importance of leveraging this information to deliver the right drug to the right patient at the right time. Key challenges in learning predictive models for this include the high-dimensionality of omics data and heterogeneity in biological and clinical factors affecting patient response. The use of multi-task learning techniques has been widely explored to address dataset limitations for in vitro drug response models, while domain adaptation (DA) has been employed to extend them to predict in vivo response. In both of these transfer learning settings, noisy data for some tasks (or domains) can substantially reduce the performance for others compared to single-task (domain) learners, i.e. lead to negative transfer (NT). RESULTS: We describe a novel multi-task unsupervised DA method (TUGDA) that addresses these limitations in a unified framework by quantifying uncertainty in predictors and weighting their influence on shared feature representations. TUGDA's ability to rely more on predictors with low-uncertainty allowed it to notably reduce cases of NT for in vitro models (94% overall) compared to state-of-the-art methods. For DA to in vivo settings, TUGDA improved over previous methods for patient-derived xenografts (9 out of 14 drugs) as well as patient datasets (significant associations in 9 out of 22 drugs). TUGDA's ability to avoid NT thus provides a key capability as we try to integrate diverse drug-response datasets to build consistent predictive models with in vivo utility. AVAILABILITYAND IMPLEMENTATION: https://github.com/CSB5/TUGDA. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.