Biology subjects

Sidi, T.

Publications and source records attributed to Sidi, T..

4 recordsLinked to original sources

bioRxiv2025

Fast and accurate taxonomic domain assignment of short metagenomic reads using BBERT

Shotgun metagenomes from complex environments such as soil uncover vast biodiversity. Yet most short reads produced by shotgun sequencing cannot be taxonomically or functionally annotated, as they lack a sufficiently comprehensive reference, obscuring the true structure and function of microbial communities. We introduce BBERT, a nucleotide large language model optimized for short reads. Testing on a large cohort of soil metagenomes, we found that BBERT identifies bacterial sequence syntax without relying on reference databases, enabling accurate assignment of taxonomic domain, coding potential, and reading frame directly from reads as short as 100 bp. BBERT is small and fast enough to analyze metagenomes using a modest GPU and can be used to convert short metagenomic reads directly to bacterial amino acid sequences for downstream applications. BBERT also improves de-novo metagenomic assembly, reducing mismatches and gaps while accelerating runtime. Using metagenomes from wild legume nodules, we demonstrate that BBERT filtering improves bin quality while significantly accelerating de-novo assembly. By providing fast, reference-free classification of short reads, BBERT unlocks large metagenomic archives for more accurate ecological and evolutionary analyses.

bioinformatics↗

bioRxiv2024

Predicting gene sequences with AI to study evolutionarily selected codon usage patterns

Selective pressure acts on the codon use, optimizing multiple, overlapping signals that are only partially understood. We trained artificial intelligence (AI) models to predict the codons given their amino acid sequence in the eukaryotes Saccharomyces cerevisiae and Schizosaccharomyces pombe and the bacteria Escherichia coli and Bacillus subtilis, to study the extent to which we can learn patterns in naturally occurring codons to improve predictions. We trained our models on a subset of the proteins, and evaluated their predictions on large, separate sets of proteins of varying lengths and expression levels. Our models significantly outperformed naive frequency-based approaches, demonstrating that there are dependencies between codons that can be learned to better predict evolutionary-selected codon usage. The prediction accuracy advantage of our models is greater for highly expressed genes and it is greater in bacteria than eukaryotes, supporting the hypothesis that there is a monotonic relationship between selective pressure for complex codon patterns and effective population size. Also, in S. cerevisiae and bacteria, our models were more accurate for longer proteins, suggesting that the AI system may have learned patterns related to co-translational folding. Gene functionality and conservation were also important determinants that affect the performance of our models. Finally, we showed that using information encoded in homologous proteins has only a minor effect on prediction accuracy, perhaps due to complex codon-usage codes in genes undergoing rapid evolution. In summary, our study employing contemporary AI methods offers a new perspective on codon usage patterns and a novel tool to optimize codon usage in endogenous and heterologous proteins. Significance statementCan one predict codon sequences used by an organism to encode a given amino acid sequence? This is difficult, because there are exponentially many codon sequences that can encode the same amino acid sequence and evolution is stochastic. Indeed, codons frequencies vary, a phenomenon known as codon-bias, yet we improve upon frequency-based predictions using contemporary AI tools that learn complex patterns and capture interactions between codons. Because our predictions are tested fairly, on cases not seen during the training process, accurate predictions suggest that these learned patterns are not random, and may be related to the evolutionary process. Thus, studying where our predictions are more accurate, is expected to reveal novel insights related to the way evolution shapes coding regions.

evolutionary biology↗

bioRxiv2019

Loss-functions matter, on optimizing score functions for the estimation of protein models accuracy

MotivationMethods for protein structure prediction (PSP) generate multiple alternative structural models (aka decoys). Thus, supervised learning methods for the evaluation and ranking of these models are crucial elements of PSP. Supervised learning involves optimization of loss functions, but their influence on performance is typically overlooked. Here we put the loss functions in the spotlight, and study their effect on prediction performance.\n\nResultsHere we report the performances of three variants of MESHI-score, a supervised learning method for the estimation of model accuracy (EMA). Each variant was trained with a different loss function and showed better performance in different aspects of the EMA problem. Most importantly, better discrimination between models of the same target, is gained by target centered loss functions.\n\nAvailabilityAll data is available at http://meshi1.cs.bgu.ac.il/SidiAndKeasar2018Data_download/. The MESHI-package (version 9.412) is available at https://github.com/meshiprot/meshi/releases).\n\nContactchen.keasar@gmail.com

bioinformatics↗

bioRxiv2017

Methods For Estimation Of Model Accuracy In CASP12

Methods for reliably estimating the quality of 3D models of proteins are essential drivers for the wide adoption and serious acceptance of protein structure predictions by life scientists. In this paper, the most successful groups in CASP12 describe their latest methods for Estimates of Model Accuracy (EMA). We show that pure single model accuracy estimation methods has shown clear progress since CASP11; the three top methods (MESHI, ProQ3, SVMQA) all perform better than the top method of CASP11 (ProQ2). The pure single model accuracy estimation methods outperform quasi-single (ModFOLD6 variations) and consensus methods (Pcons, ModFOLDclust2, Pcomb-domain and Wallner) in model selection, but are still not as good as those methods in absolute model quality estimation and predictions of local quality. Finally, we show that when using contact based model quality measures (CAD, 1DDT) the single model quality methods perform relatively better.

bioinformatics↗