Search bioRxiv⌕ Search

Biology subjects

Raya, D.

Publications and source records attributed to Raya, D..

2 recordsLinked to original sources

Predicting cyanobacterial oxic diazotrophy and prioritizing candidate protein families beyond known Nif proteins

Diazotrophic cyanobacteria must reconcile oxygenic photosynthesis with the oxygen sensitivity of nitrogenase. Engineering studies have demonstrated nitrogenase activity in a non-diazotrophic cyanobacterium while highlighting the importance of accessory functions and oxygen management. Robust aerobic nitrogen fixation remains challenging, motivating systematic identification of candidate support functions beyond the canonical Nif machinery. We modeled nifHDK-defined diazotrophic potential from protein-family presence or absence in 426 cyanobacterial assemblies. Four classifiers evaluated with genus-blocked cross-validation achieved mean ROC AUCs of 0.853 to 0.893, with purity and genus-breadth filters fitted within training folds. Restricting the panel to 401 assemblies with current NCBI status gave AUCs of 0.806 to 0.881. Family-content logistic regression outperformed a genome-size-only baseline across ten genus partitions. Two inventories describe candidates across the full panel: 981 high-purity families and 473 families with positive residual associations (accompanied by three ubiquitous controls). Model importance orders the residual-association inventory. A separate heuristic score combines inventory membership with proteomics, condensate-driver ranks and genus-based morphotype proxies to prioritize 167 families for experiments. Proteomics annotations tracked detection opportunity; the condensate contrast between inventories attenuated after prevalence adjustment. These annotations help choose experiments but do not independently validate accessory functions. Protein-family content predicts nifHDK-defined diazotrophic potential, with performance sensitive to assembly inclusion, taxonomy, genome size and taxonomic weighting. The two candidate inventories distinguish high carrier purity from residual association, providing complementary resources for prioritizing previously uncharacterized functions associated with cyanobacterial oxic diazotrophy.

bioinformatics↗

Classification of bacterial nanowire proteins using Machine Learning and Feature Engineering model

Nanowires (NW) have been extensively studied for Shewanella spp. and Geobacter spp. and are mostly produced by Type IV pili or multiheme c-type cytochrome. Electron transfer via NW is the most studied mechanism in microbially induced corrosion, with recent interest in application in bioelectronics and biosensor. In this study, a machine learning (ML) based tool was developed to classify NW proteins. A manually curated 999 protein collection was developed as an NW protein dataset. Gene ontology analysis of the dataset revealed microbial NW is part of membranal proteins with metal ion binding motifs and plays a central role in electron transfer activity. Random Forest (RF), support vector machine (SVM), and extreme gradient boost (XGBoost) models were implemented in the prediction model and were observed to identify target proteins based on functional, structural, and physicochemical properties with 89.33%, 95.6%, and 99.99% accuracy. Dipetide amino acid composition, transition, and distribution protein features of NW are key important features aiding in the models high performance.

systems biology↗