Fig. 4
From: HMD-ARG: hierarchical multi-task deep learning for annotating antibiotic resistance genes
Functional validation of the predicted ARGs and structural investigation of conserved sites detected by HMD-ARG. a Left figure: A diagram showing the procedure of heterologous expression and functional analysis of the predicted ARGs from PA150567 in E. coli BL21 host. Middle figure: Growth curves of E. coli hosts with expression of the predicted β-lactamases that inactivate β-lactam antibiotics in the presence of 2 μg/ml meropenem. Right figure: Growth curves of E. coli hosts with expression of the predicted acetyltransferases that inactivate aminoglycoside antibiotics in the presence of 4 μg/ml amikacin. b The HMD-ARG predicted conserved sites and the corresponding sequence logo from 319 to 393 in AXX01_04100. c After we mutated the conserved site (346) from T to N, the mutated protein’s (colored in red) local structure (predicted by RaptorX) changed significantly from the wild type (colored in light blue). Moreover, the binding affinity (predicted by AutoDock) between the mutated protein and the ligand (antibiotics) also reduced, as illustrated in the middle (wild type) and right (mutated protein) figures