Journal of Natural Products

The incidence of infections caused by multidrug-resistant bacterial agents has increased at an alarming rate worldwide. With the aim of identifying novel antimicrobial peptides (AMPs), we screened bacteria isolated from various environmental samples. Our hypothesis was that the sequence space of natural AMPs belonging to known AMP classes is far from saturation. We used a classical pipeline of bacterial culturing, overlay assays, extraction and fractionation to purify AMPs for identification. Using LC/MS analysis, the structure of AMPs isolated from a strain of Paenibacillus was narrowed down and a new subclass of polymyxins was uncovered. These harbor three aliphatic residues within the cyclic C-terminal, and in certain cases Ser replaces Thr at position A2. Four discrete polymyxins fitting the new class, but not identical to the natural polymyxins were synthesized and tested against 29 human pathogenic bacteria. Each displayed an antibacterial spectrum different from that of colistin, with the best candidate surpassing it in potency against ten bacterial strains, but underperforming against three others. These results demonstrate that current screening of natural bacterial isolates can still permit the design of novel antimicrobial peptide variants without the substantial threat of diminishing returns.

The 4-amidopentadienoate (APD) motif is a rare and bioactive pharmacophore found in the cyclic lipodepsipeptide natural products BE-43547, rakicidins, microtermolides, and related compounds. Although several biosynthetic gene clusters associated with this family have been identified, the enzymatic origin of the APD warhead has remained unknown. Here, we investigated APD formation using NRPS engineering to install the unusual APD-associated polyketide synthase (PKS) module from the BE-43547 biosynthetic pathway into heterologous nonribosomal peptide synthetase/polyketide synthase (NRPS/PKS) assembly lines expressed in Escherichia coli. The engineered hybrids produced the designed lipopeptide containing the APD moiety, demonstrating that the core biosynthetic machinery is sufficient for APD installation and that no trans-acting enzymes are required. Total synthesis of the target compound enabled structural validation by co-elution, MS/MS analysis, and NMR spectroscopy. Production titers of 59 ug/L were achieved in the heterologous host. Systematic mutagenesis further revealed that the conserved dehydratase His-Asp catalytic dyad is essential for APD formation, whereas an additional highly conserved histidine is dispensable. These findings establish the biosynthetic origin of the APD warhead and provide a foundation for incorporating this privileged pharmacophore into engineered peptide scaffolds.

Endophytic fungi associated with the Himalayan yew (Taxus wallichiana) represent an underexplored source of bioactive secondary metabolites. This study investigated the extracellular metabolites of Annulohypoxylon purpureonitens isolated from Nepalese T. wallichiana using bioactivity screening combined with LCMS/MS-based metabolomics. The fungal extract exhibited broad-spectrum antibacterial activity, showing the strongest inhibition against Staphylococcus aureusand Enterococcus faecalis (MIC = 500 ug/mL). It also displayed notable antioxidant capacity(DPPH, ABTS, TPC &TFC) and cytotoxicity against HeLa and MCF-7 cancer cell lines. Metabolite profiling via GNPS molecular networking, manual MS/MS validation, and MASST reverse metabolomics putatively identified diverse compounds, including hydroquinidine, chlorogenic acid, muramic acid, and cordycepin conjugates widely distributed across public microbial datasets. Overall, A. purpureonitens is a promising source of multifunctional metabolites, laying a foundation for future compound isolation and functional characterization.

Oxalyl-coenzyme A (oxalyl-CoA) is a key intermediate in oxalate metabolism in plants, fungi and oxalate-degrading bacteria, but its limited availability has restricted biochemical investigations of oxalyl-CoA-dependent enzymes. Here, we describe a practical semisynthetic procedure for the preparation of oxalyl-CoA based on rapid oxalyl transfer from S-oxalyl p-thiocresol to coenzyme A. The reaction was monitored directly by 1H NMR spectroscopy, allowing optimisation of pD and reaction conditions. Following removal of thiocresol and purification by reversed-phase HPLC, oxalyl-CoA was obtained in 39% yield as determined by quantitative 1H NMR. The product was characterised by high-resolution electrospray mass spectrometry and comprehensive 1H, 13C and 31P NMR spectroscopy, confirming its structure unequivocally. During the study, the limited stability of oxalyl-CoA in aqueous solution was documented, leading to recommendations for its purification and storage. The semisynthetic protocol provides a convenient source of analytically pure oxalyl-CoA suitable for biochemical assays and supplies reference spectroscopic data for its unambiguous identification. The biological utility of the semisynthetic oxalyl-CoA was demonstrated by its application as an acyl donor substrate in assays of PnBAHD15, enabling quantitative kinetic characterisation of the enzyme and illustrating its suitability for biochemical studies of oxalyl-CoA-dependent enzymes.

Lanthipeptides are one of the largest classes of ribosomally synthesized and post-translationally modified peptides (RiPPs). The coi biosynthetic gene cluster (BGC) from Streptomyces coelicolor A3(2) encodes a canonical class I lanthipeptide dehydratase (CoiB) and cyclase (CoiC), a bifunctional enzyme (CoiSA) with an O-methyltransferase (MT) and glutamyl lyase (GL) domain, and a protein of unknown function (CoiH). The product of the coi BGC was recently shown to impart anti-phage activity, but its structure is still unresolved. Previous work investigated the regioselectivity of the GL domains in CoiB and CoiSA and the stereochemistry of the cyclized precursor peptide, but the function of CoiH was not addressed. In this study, co-expression of the peptide CoiA1 with CoiBCSAH resulted in a +16 Da addition on the cyclized peptide compared to when CoiH was omitted. LC-MS/MS analysis indicated that this modification occurred in the first thioether ring. A combination of site-directed mutagenesis, comparison of linear and cyclized peptide substrates, hydrogen peroxide (H2O2) treatment, and collision-induced dissociation (CID) mass spectrometric analysis suggested that the sulfur atom in the first methyllanthionine was oxidized to a sulfoxide group by CoiH. This hypothesis was confirmed by NMR analysis. CoiH represents a previously uncharacterized oxygenase family catalyzing sulfoxide formation. Structure prediction tools suggest a novel enzyme fold without obvious metal or cofactor binding sites, raising the possibility that CoiH is a cofactor independent oxidation enzyme.

CC10137 is a first-in-class peripherally restricted 2A-adrenergic receptor (2AAR) agonist with broad-spectrum analgesic efficacy and a favorable safety profile. Systematic exploration of the chemical space surrounding first-in-class leads is important for defining series boundaries and guiding continued optimization, but conventional analogue-by-analogue medicinal chemistry samples only a small fraction of the accessible structural space. Here, we used a scaffold-seeded enumeration strategy to expand the chemical space surrounding CC10137 from four SAR-informed seed compounds comprising CC10137 and three closely related structural variants. Application of predefined medicinal chemistry transformation rules in StarDrop generated a virtual library of 16,601,163 unique structures. Morgan fingerprint-based principal component analysis indicated that the library occupied a highly multidimensional structural space involving variation in scaffold substitution, peripheral functional groups, and side-chain composition. A retrospective comparison set of 43 compounds independently designed and experimentally characterized in the earlier CC10137 program represented only approximately 0.00026% of the 16.6-million-member library, yet all 43 were recovered as exact structural matches. Three compounds selected directly from the virtual library retained 2AAR binding affinity and agonist potency below 25 nM. Five representative compounds further showed significant anti-allodynic effects in the in vivo spared nerve injury model, with inhibition rates ranging from 39.3% to 55.7%. These findings support scaffold-seeded computational enumeration as a practical strategy for systematic chemical-space mapping around a first-in-class lead and for identifying additional pharmacologically active structural regions for further optimization.

Tandem mass spectrometry has become central to untargeted metabolomics. The translation of unknown spectra into biological insight depends on assigning chemical identities to detected metabolites. Structural characterization typically begins with mass spectral library matching, in which experimental spectra are compared against reference libraries and candidate annotations are ranked by their spectral similarity to the query. As spectral libraries and experimental datasets grow, however, more candidates achieve comparable similarity scores for a single query, and similarity scores give no indication of how reproducible a candidate match is or how sensitive it is to the underlying fragment evidence. Existing false-discovery-rate approaches can indicate annotation error at the dataset level but do not provide a per-match estimate of reliability. Here, we introduce a SpecReBoot-inspired query-focused bootstrapping approach that resamples the fragment evidence of each query spectrum. This approach relies on recomputing query similarity to candidate library spectra across bootstrap replicates, which provides a statistical distribution of scores rather than a single value. From this distribution we define the match support, a per-match reliability estimate quantifying the reproducibility of a match under spectral perturbation, together with measures of ranking stability that describe how often a candidate remains among the top-ranked matches across replicates. Applied to a forensic drug-of-abuse case, match support distinguished previously identified annotations from high-scoring false positives: a distinction cosine similarity failed to make. Furthermore, match support values remained stable as the reference library was expanded, whereas ranking stability metrics shifted significantly. In a cross-instrument endogenous metabolite library search, match support further revealed metric-specific annotation behavior, identifying metabolites consistently supported across different similarity metrics, while flagging annotations whose reliability depended strongly on the chosen scoring metric. Benchmarking against a natural-product reference library demonstrated that ranking based on match support values promoted true matches by four ranks on average compared with cosine-based ranking, without promoting analogs. Under controlled spectral perturbation experiments, match support flagged incorrect annotations with an AUROC of 0.75, whereas the cosine similarity score alone of the same match reached only 0.56. Query-focused bootstrapping thus provides a practical, per-match measure of annotation reliability, bringing the field a step toward reliable annotations at scale. We anticipate that incorporation of our annotation reliability scoring into computational metabolomics workflows will further promote the growth of spectral libraries and enhance their applicability across scientific disciplines.

Marine macroalgae host endophytic microbial communities that serve as valuable reservoirs for drug discovery. In this study, the endophytic actinomycete Streptomyces sp. UL09 was isolated from the internal thallus tissues of the green macroalga Ulva rigida C. Agardh. Fermentation of strain UL09 yielded the nitrobenzoyl sesquiterpenoids insulicolide B (1.63 mg/L) and insulicolide C (1.79 mg/L). The planar structures of compounds 1 and 2 were assigned as insulicolides B and C, respectively, via spectroscopic analysis and comparison with literature data. In vitro bioassays revealed moderate activity against Gram-positive bacteria (MIC 128-512 g/mL), including MRSA Sp3 (MIC 128-256 g/mL). These compounds exhibited cytotoxic activity against human cancer cell lines (HeLa, HepG2, and MDA-MB-231) with IC50 values ranging from 34.35 to 142.64 g/mL. Both compounds demonstrated lower cytotoxicity toward non-malignant Vero cells. Molecular docking simulations predicted that both metabolites target the catalytic ATP-binding cleft of human Polo-like kinase 1 (Plk1), with insulicolide B displaying a favorable docking score ({Delta}G = -9.273 kcal/mol). Furthermore, pkCSM predicted high intestinal absorption and no alerts for the assessed toxicity endpoints. These findings establish Streptomyces sp. UL09 as a viable alternative biological source of insulicolides B and C, highlighting their potential as leads for oncology applications.

Screening of the open-access CRESTdb small-molecule library identified the imidazothiazole carboxamide sCQG200 (1) as an initial hit, with activity against H37Rv and Erdman Mycobacterium tuberculosis strains in cholesterol-containing medium (MIC90= 6.88 uM and 4.48 uM, respectively) and against intracellular Mtb (EC50= 3.80 uM). sCQG200 retained activity against drug-resistant Mtb strains, including clinical MDR/XDR isolates. SAR optimization led to analogs 21 and 24 with substantially improved antimycobacterial potency. Mouse pharmacokinetic studies following oral and intravenous administration showed 62.1% and 29.5% oral bioavailability for 21 and 24, respectively. This structurally differentiated chemotype provides a promising starting point for further optimization as antitubercular agent.

Rv1625c/Cya has emerged as a promising target for the development of treatment-shortening therapies for tuberculosis. Screening of an Enamine compound library identified sBQQ004 as an initial hit, and rapid hit optimization led to compound 1, which was subsequently confirmed as an Rv1625c/Cya agonist. Structure activity relationship studies identified lead compound 25 with potent activity against Mycobacterium tuberculosis H37Rv under cholesterol-dependent growth conditions (MIC = 0.27 uM) and strong intramacrophage activity (EC50 = 0.079 uM). Compounds 1 and 25 showed oral bioavailabilities of 84.5% and 51.9% in mice, respectively. Repeat BID dosing of compound 1 resulted in a dose- and time-dependent decrease in systemic exposure. Despite this pharmacokinetic limitation, the potency and overall profile of this chemotype encouraged continued optimization.

Mycosporine-like amino acids (MAAs) are among nature's most effective sunscreen compounds, capable of converting harmful ultraviolet radiation into harmless heat, and are widely distributed in marine organisms such as red macroalgae. Although decades of research have led to numerous discoveries, the rate of new MAA identifications has declined. To address this, we considerably expanded our previously developed combinatorial MAA database, increasing the number of covered structures tenfold. Following a comprehensive literature search for plausible but undescribed building blocks, the database now incorporates an extensive set of proteinogenic and non-proteinogenic amino acids, as well as other marine organic osmolytes, in combination with all (currently) known MAA scaffolds. This expanded resource was integrated into our identification platform, which combines UHPLC-VWD-HRMS2 analysis, feature-based molecular networking, and bioinformatics-driven annotation. Application of this updated workflow enabled the isolation and structural elucidation of a novel MAA, mycosporine-cysteinolic acid, from the red marine macroalga Vertebrata lanosa. Altogether, this study provides a valuable extension of the bioinformatics-based MAA screening pipeline, enhancing the annotation and discovery of novel MAAs in natural matrices.

Arylamino alcohols constitute an established class of antimalarial agents. While this class is best known for its potent antiplasmodial activity, selected members, most notably mefloquine, have also been reported to exhibit antibacterial activity. Structurally related 3-hydroxypropanamidines were developed as highly potent antimalarial agents. Here, we explored the antibacterial activity of three 3-hydroxypropanamidines (4 (BLK278), 5 (BLK280), and 6 (TKK088)) against clinically relevant Gram-positive and Gram-negative bacteria and investigated their cytotoxicity, in vitro pharmacokinetic properties, and proteome-based bacterial response to treatment. We observed good antibacterial activity against Staphylococcus aureus, with 6 (TKK088) displaying the highest activity against multidrug-resistant MRSA and VISA strains. 5 (BLK280) demonstrated the greatest spectrum of activity acting also against clinically relevant Gram-negative strains of Escherichia coli, Acinetobacter baumannii, Klebsiella pneumoniae, Klebsiella aerogenes, and Shigella sonnei. In cytotoxicity assays with HUVEC and HAOSMC cells, 4 (BLK278) and the previously reported 7 (TKK129) showed the most favourable overall profiles among the investigated 3-hydroxypropanamidines. The selectivity indices for S. aureus were below 2.5 with clear cell-type-dependent differences for all compounds including mefloquine, suggesting that improving antibacterial selectivity is an important objective for further optimization. In vitro pharmacokinetics showed high plasma protein binding and good human microsomal and plasma stability for all three 3-hydroxypropanamidines. Gel-based proteomic responses revealed a strong impact on the bacterial cell envelope. The observed acute proteomic responses reflect pleiotropic effects and are consistent with earlier reports implicating F0F1-ATPase as a potential molecular target of mefloquine in Streptococcus pneumoniae.

Antibiotics with new mechanisms are highly pursued to address the threat of infections caused by drug-resistant Gram-negative bacteria. Targeting MsbA, a key protein of the lipopolysaccharide biosynthesis pathway, represents a promising strategy to discover new classes of antibiotics. However, currently available MsbA-targeted molecules either lack sufficient potency or have unfavorable properties, necessitating expansion of chemical space. In this study, we chose the most promising cerastecin Cpd 4 as the template, and used two Artificial Intelligence (AI)-based tools, i.e. Link-INVENT and AutoMolDesigner for molecular design, performed chemical derivatization and antibacterial activity evaluation, which led to the discovery of Y-11 (MIC for A. baumannii: 0.5 g/mL). Encouragingly, Y-11 showed equivalent potency to Cpd4 for carbapenem-resistant A. baumannii, and less cytotoxicity and hemolysis as well as lower spontaneous resistance frequency. In vivo efficacy study demonstrated that Y-11 could effectively reduce bacterial loads in the mice infected by A. baumannii. The following mechanism study including molecular dynamics simulation, biochemical assay, and transmission electron microscope (TEM) analysis suggested that Y-11 inhibited the transport of lipooligosaccharide and impaired the formation of outer membrane, probably by competitively binding to the substrate binding site of MsbA and modulating ATPase activity. Taken together, we have discovered a MsbA-targeted small molecule Y-11 via AI-driven drug design, which provides a foundation for future antibiotic development.

Iron is a key micronutrient that constrains microbial growth and productivity in many aquatic and terrestrial environments due to its limited bioavailability. Microorganisms evolved sophisticated acquisition strategies, including the production of siderophores, high-affinity iron-chelating molecules that facilitate iron solubilisation and uptake. Cyanobacteria, photosynthetic prokaryotes and major contributors to global primary production, also depend on iron as a cofactor to their core metabolic enzymes. However, very few cyanobacterial siderophores were described so far, and cyanobacteria remain an underxplored source of possibly novel siderophores. Here we report a novel cyanobacterial siderophore, cyanochelin C, that employs two {beta}-hydroxyaspartate residues for iron chelation. We provide extensive nuclear magnetic resonance (NMR) and mass spectrometry (MS) evidence on the molecular structure and identify the corresponding biosynthetic gene cluster (BGC). Bioinformatic analysis of the BGC further revealed the presence of an acylase CcsQ clustering with a broader cyanobacteria-specific family of acylases associated with predicted siderophore-encoding BGCs. Discovery of cyanochelin C and its deacylation by CcsQ expands the known structural diversity of cyanobacterial siderophores and improves the understanding of important enzymatic reactions.

Nature produces hundreds of carotenoids, yet only a handful of the apocarotenoids derived from them are accessible through microbial production. The best-known example is retinal, the chromophore of rhodopsins and a precursor of pharmaceutical retinoids, which is generated by the central cleavage of {beta}-carotene. Whether the same cleavage chemistry can be extended to other carotenoids, yielding retinal analogues that differ in their ring structures, and potentially in their biological activities, has remained largely untested. In this study, we demonstrate a pathway engineering approach in E. coli for the biosynthesis of diverse retinal analogues by leveraging substrate promiscuity of Blh, a bacterial carotenoid cleavage enzyme originally identified in microbial rhodopsin gene clusters. While initial co-expression of Blh with carotenoid pathway genes often resulted in the production of retinal (by cleavage of {beta}-carotene intermediate), we found that by optimizing the expression level of Blh, carotenoids such as astaxanthin or canthaxanthin were cleaved efficiently. Structure-guided engineering of Blh, informed by its predicted substrate-binding cavity, further improved the cleavage of zeaxanthin. This expanded catalytic activity suggests that Blh can serve as a versatile biocatalyst for the production of diverse retinal analogues, potentially yielding compounds with a range of biological activities. Furthermore, our findings raise the possibility of diverse biological roles for these enzymes in their native biological contexts. ImportanceThis study demonstrated the successful biosynthesis of a diverse array of retinal analogues in engineered Escherichia coli through the heterologous expression of Blh, a {beta}-carotene cleavage dioxygenase, together with several carotenoid pathways. Careful design of the Blh expression construct enabled modulation of retinoid proportions in the engineered pathway. This work uncovers previously unrecognized substrate promiscuity of Blh, revealing its capacity to accept carotenoids beyond {beta}-carotene as substrates. For the first time, the predicted structure of Blh revealed the enzymes substrate cavity. Rational engineering by amino acid substitution designed to expand the cavity enabled the improved cleavage of hydroxylated carotenoids. These findings open new avenues for both fundamental research and biotechnological applications and have the potential to impact the microbial production of valuable retinoids.

Selective peptide agonists for complement C5a receptor 1 (C5aR1) are valuable tools for dissecting receptor-specific inflammatory signalling, but their optimisation is complicated by overlap with closely related anaphylatoxin receptors and by pathway-dependent pharmacology. Here, we applied a structure-guided computational workflow to prioritise mutations within two C5a-derived peptide agonist scaffolds. FoldX-guided modelling identified position 5 as a candidate optimisation site, with hydrophobic substitutions predicted to improve C5aR1 engagement without corresponding gains at C3aR. Predicted BM1 and BM221 analogues were synthesised by solid-phase peptide synthesis and evaluated across C3aR, C5aR1 and C5aR2 using ERK1/2 phosphorylation and {beta}-arrestin recruitment assays. Position-5 substitutions enhanced C5aR1 functional preference in ERK assays, although additional replacement of Leu6 with Ala reduced target potency and revealed pathway-dependent receptor discrimination. BM1 P5M provided the clearest overall improvement across ERK and {beta}-arrestin readouts. In the BM221 series, A5Nle improved C5aR1 preference over C3aR, whereas A5Nle Abu6Ala produced the most favourable serum stability profile. These findings support position 5 as a transferable optimisation site and demonstrate that C5aR1 potency and receptor selectivity must be balanced during next-generation agonist design.

The classical estrogen receptor (ER) is a steroid hormone receptor known for its role in both health and disease. ER, like its counterpart ER{beta}, acts as a receptor for its endogenous ligand 17{beta}-estradiol (E2) and is capable of binding a diverse set of natural and synthetic small molecules. Several screening campaigns have identified various ER ligands, based on the oxabicyclo[3.3.1]-nonene scaffold, that act as ER agonists. One such campaign extensively characterized an oxabicyclic ER agonist, termed AB-1. Herein, we report the synthesis of a series of analogs based on the scaffold of AB-1 (herein termed 1), focusing on the enhancement of their binding to ER and functional potencies in ER mediated assays. We identify a new substituted compound 9 that achieves subnanomolar potencies for ER binding (IC50= 0.45 nM) and ER dependent transcription (EC50= 0.37 nM) and proliferation (EC50= 0.03 nM), with potencies approaching those of E2. The high potency of 9 provides a promising scaffold for the potential development of new ER-targeted therapeutics.

GPR183 is an oxysterol-sensing GPCR predominantly expressed in lymphoid organs and tissues. Activation of the receptor by oxysterol 7,25-OHC leads to Gi protein-mediated signaling as well as {beta}-arrestin2 recruitment. GPR183/oxysterol signaling modulates localization of lymphoid cells, consequently the receptor is associated with several inflammation-associated diseases and is an interesting potential drug target. Previously, we reported the discovery of moderately potent G protein-biased partial agonists for GPR183 from a virtual screening based on the scaffold of the antagonist NIBR189. Herein, we present the detailed structure-activity investigations and optimizations, which led to the identification of full agonists for GPR183 with complete bias for Gi protein signaling and low nanomolar potency, including 63 (TUG-2604) with potency and efficacy similar to 7,25-OHC. Notably, 63 was unable to induce migration of human dendritic cells but inhibited migration induced by 7,25-OHC. This compound will be valuable for further explorations of the signaling-specific function and drug target potential of GPR183.

The fosfazinomycins are phosphonate natural products with antifungal activity. Their main scaffold is composed of a phosphonate moiety attached to the carboxylate of arginine via a hydrazine linkage. Previous studies have elucidated a convergent biosynthetic pathway that independently assembles phosphonate and hydrazine synthons, but how these building blocks are then connected could not be determined. In this work, we provide the final missing steps in fosfazinomycin biosynthesis by revealing an unexpected engagement of biosynthetic machinery that is typically involved in ribosomally synthesized and post-translationally modified peptides (RiPPs). An asparagine synthetase-like (AS-like) enzyme catalyzes the installation of hydrazine onto the carboxylate of Arg at the C-terminus of a short ribosomally synthesized precursor peptide. The terminal nitrogen of the resulting peptide hydrazide is methylated, and the phosphonate moiety is activated to a triphosphate-like intermediate through two separate kinase catalyzed phosphorylation steps. A nucleotidyl transferase then catalyzes the ligation of the two fragments to generate the mature fosfazinomycin scaffold on a peptide. Aminopeptidase cleavage of this peptide then yields fosfazinomycin B (fosB), which serves as substrate for a valinyl-tRNA dependent reaction to afford fosfazinomycin A (fosA). This work demonstrates an unprecedented example of the convergence of RiPP and phosphonate biosynthetic logic, an enzymatic route to peptide hydrazides that are widely used in peptide ligation chemistry, and an unusual activation sequence for conjugation of phosphonates.

As a consequence of the overuse of conventional antibiotics, there is currently an unprecedented increase in antibiotic resistance in newer generations of pathogenic bacteria. This growing problem has led scientists to discover novel medications that could potentially reduce the usage of antibiotics, such as antimicrobial peptides (AMPs). Recent study has demonstrated that pardaxin could bind deeply into the surface of a lipid model membrane and inhibit the growth of pathogenic bacteria, including Staphylococcus aureus and Escherichia coli. In this work, the antibacterial efficacy of pardaxin was extended further by performing selective in silico substitution, driven by deep learning, Evolutionary-Scale Cambrian (ESMC) combined with conventional AMP design principles. Principal component analyses of the ESMC embeddings combined with conventional design models produced a set of single-and multiple-mutant analogues of pardaxin, which were predicted and validated with experimental data to exhibit selective antimicrobial properties against either E. coli or S. aureus, respectively. Atomistic molecular dynamics simulations further supported the notion that alpha-helical stability is a critical predictor of inner membrane activity, strongly correlated with their selective antimicrobial action tested in the lab. Overall, the findings highlighted a promising application of deep evolutionary machine learning techniques for screening a range of novel AMPs for selective antimicrobial agents.