Task-Similarity is a Crucial Factor for Few-Shot Meta-Learning of Structure-Activity Relationships.

Machine learning models support computer-aided molecular design and compound optimization. However, the initial phases of drug discovery often face a scarcity of training data for these models. Meta-learning has emerged as a potentially promising strategy, harnessing the wealth of structure-activity data available for known targets to facilitate efficient few-shot model training for the specific target of interest. In this study, we assessed the effectiveness of two different meta-learning methods, namely model-agnostic meta-learning (MAML) and adaptive deep kernel fitting (ADKF), specifically in the regression setting. We investigated how factors such as dataset size and the similarity of training tasks impact predictability. The results indicate that ADKF significantly outperformed both MAML and a single-task baseline model on the inhibition data. However, the performance of ADKF varied across different test tasks. Our findings suggest that considerable enhancements in performance can be anticipated primarily when the task of interest is similar to the tasks incorporated in the meta-learning process.

Morphing of Amphipathic Helices to Explore the Activity and Selectivity of Membranolytic Antimicrobial Peptides.

Naturally occurring membranolytic antimicrobial peptides (AMPs) are rarely cell-type selective and highly potent at the same time. Template-based peptide design can be used to generate AMPs with improved properties de novo. Following this approach, 18 linear peptides were obtained by computationally morphing the natural AMP Aurein 2.2d2 GLFDIVKKVVGALG into the synthetic model AMP KLLKLLKKLLKLLK. Eleven of the 18 chimeric designs inhibited the growth of Staphylococcus aureus, and six peptides were tested and found to be active against one resistant pathogenic strain or more. One of the peptides was broadly active against bacterial and fungal pathogens without exhibiting toxicity to certain human cell lines. Solution nuclear magnetic resonance and molecular dynamics simulation suggested an oblique-oriented membrane insertion mechanism of this helical de novo peptide. Temperature-resolved circular dichroism spectroscopy pointed to conformational flexibility as an essential feature of cell-type selective AMPs.

Identification of Chemokine Ligands by Biochemical Fragmentation and Simulated Peptide Evolution.

Short linear peptides can overcome certain limitations of small molecules for targeting protein-protein interactions (PPIs). Herein, the interaction between the human chemokine CCL19 with chemokine receptor CCR7 was investigated to obtain receptor-derived CCL19-binding peptides. After identifying a linear binding site of CCR7, five hexapeptides binding to CCL19 in the low micromolar to nanomolar range were designed, guided by pharmacophore and lipophilicity screening of computationally generated peptide libraries. The results corroborate the applicability of the computational approach and the chosen selection criteria to obtain short linear peptides mimicking a protein-protein interaction site.

Simulated Molecular Evolution for Anticancer Peptide Design.

A computational technique based on a simulated molecular evolution protocol was employed for anticancer peptide (ACP) design. Starting from known ACPs, innovative bioactive peptides were automatically generated in computer-assisted design-synthesize-test cycles. This design algorithm offers a viable strategy for the generation of novel peptide sequences, without requiring a priori structure-activity knowledge. Sequence morphing and activity improvement were achieved through iterative amino acid variation and selection. Results show that not only the interaction of ACPs with the target membrane is important for their anticancer activity, but also the degree of peptide dimerization, which was corroborated by temperature profiling and electrospray mass spectrometry.

modlAMP: Python for antimicrobial peptides.

SUMMARY: We have implemented the lecular esign aboratory's nti icrobial eptides package ( ), a Python-based software package for the design, classification and visual representation of peptide data. modlAMP offers functions for molecular descriptor calculation and the retrieval of amino acid sequences from public or local sequence databases, and provides instant access to precompiled datasets for machine learning. The package also contains methods for the analysis and representation of circular dichroism spectra. AVAILABILITY AND IMPLEMENTATION: The modlAMP Python package is available under the BSD license from URL http://doi.org/10.5905/ethz-1007-72 or via pip from the Python Package Index (PyPI). CONTACT: gisbert.schneider@pharma.ethz.ch. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The N-terminal extension of the P. falciparum GBP130 signal peptide is irrelevant for signal sequence function.

The malaria parasite P. falciparum exports a large number of proteins to its host cell, the mature human erythrocyte. Although the function of the majority of these proteins is not well understood, many exported proteins appear to play a role in modification of the erythrocyte following invasion. Protein export to the erythrocyte is a secretory process that begins with entry to the endoplasmic reticulum. For most exported proteins, this step is mediated by hydrophobic signal peptides found towards the N-terminal end of proteins. The signal peptides present on P. falciparum exported proteins often differ in length from those found in other systems, and generally contain a highly extended N-terminal region. Here we have investigated the function of these extended N-terminal regions, using the exported parasite protein GBP130 as a model. Surprisingly, several deletions of the extended N-terminal regions of the GBP130 signal peptide have no effect on the ability of the signal peptide to direct a fluorescent reporter to the secretory pathway. Addition of the same N-terminal extension to a canonical signal peptide does not affect transport of either soluble or membrane proteins to their correct respective subcellular localisations. Finally, we show that extended signal peptides are able to complement canonical signal peptides in driving protein traffic to the apicoplast of the parasite, and are also functional in a mammalian cell system. Our study is the first detailed analysis of an extended P. falciparum signal peptide and suggests that N-terminal extensions of exported Plasmodium falciparum proteins are not required for entry to the secretory system, and are likely to be involved in other, so far unknown, processes.

Recurrent Neural Network Model for Constructive Peptide Design.

We present a generative long short-term memory (LSTM) recurrent neural network (RNN) for combinatorial de novo peptide design. RNN models capture patterns in sequential data and generate new data instances from the learned context. Amino acid sequences represent a suitable input for these machine-learning models. Generative models trained on peptide sequences could therefore facilitate the design of bespoke peptide libraries. We trained RNNs with LSTM units on pattern recognition of helical antimicrobial peptides and used the resulting model for de novo sequence generation. Of these sequences, 82% were predicted to be active antimicrobial peptides compared to 65% of randomly sampled sequences with the same amino acid distribution as the training set. The generated sequences also lie closer to the training data than manually designed amphipathic helices. The results of this study showcase the ability of LSTM RNNs to construct new amino acid sequences within the applicability domain of the model and motivate their prospective application to peptide and protein design without the need for the exhaustive enumeration of sequence libraries.

Quantification of hydrolyzed peptides and proteins by amino acid fluorescence.

Reliable quantification of peptides and proteins is essential for drug discovery. We report the successful development and validation of an accurate and broadly applicable high performance liquid chromatography hyphenated to fluorescence detector procedure for the quantitative determination of the aromatic amino acids tyrosine, phenylalanine, and tryptophan, without relying on derivatization chemistry. Using ion-pair chromatography, fluorescent amino acids were clearly separated within 10 minutes. The hydrolysis of peptides was performed under acidic and heated conditions to yield the monomeric building blocks. Various protecting agents were tested to ensure tryptophan stability. The presented analytical method accurately (>95%) quantifies all fluorescent residues. The power of the method was confirmed by correct quantification of protein reference standard to 98.6% over all fluorescence traces. The method allowed us to identify pre-analytical differences between the nominal and actual concentrations of 12 peptide solutions. Salt formation, weighing errors, and other pre-analytical pitfalls resulted in noteworthy differences of up to 85% between the indicated and actual concentration of peptide solutions, subsequently leading to false positive or negative interpretation of activity data. Finally, only one solution is needed to perform quantification as well as UV-purity tests and can further be used as stock solution for activity testing.

Rational Design of Membrane-Pore-Forming Peptides.

Specific interactions of peptides with lipid membranes are essential for cellular communication and constitute a central aspect of the innate host defense against pathogens. A computational method for generating innovative membrane-pore-forming peptides inspired by natural templates is presented. Peptide representation in terms of sequence- and topology-dependent hydrophobic moments is introduced. This design concept proves to be appropriate for the de novo generation of first-in-class membrane-active peptides with the anticipated mode of action. The designed peptides outperform the natural template in terms of their antibacterial activity. They form a kinked helical structure and self-assemble in the membrane by an entropy-driven mechanism to form dynamically growing pores that are dependent on the lipid composition. The results of this study demonstrate the unique potential of natural template-based peptide design for chemical biology and medicinal chemistry.

Identification of new PNEPs indicates a substantial non-PEXEL exportome and underpins common features in Plasmodium falciparum protein export.

Malaria blood stage parasites export a large number of proteins into their host erythrocyte to change it from a container of predominantly hemoglobin optimized for the transport of oxygen into a niche for parasite propagation. To understand this process, it is crucial to know which parasite proteins are exported into the host cell. This has been aided by the PEXEL/HT sequence, a five-residue motif found in many exported proteins, leading to the prediction of the exportome. However, several PEXEL/HT negative exported proteins (PNEPs) indicate that this exportome is incomplete and it remains unknown if and how many further PNEPs exist. Here we report the identification of new PNEPs in the most virulent malaria parasite Plasmodium falciparum. This includes proteins with a domain structure deviating from previously known PNEPs and indicates that PNEPs are not a rare exception. Unexpectedly, this included members of the MSP-7 related protein (MSRP) family, suggesting unanticipated functions of MSRPs. Analyzing regions mediating export of selected new PNEPs, we show that the first 20 amino acids of PNEPs without a classical N-terminal signal peptide are sufficient to promote export of a reporter, confirming the concept that this is a shared property of all PNEPs of this type. Moreover, we took advantage of newly found soluble PNEPs to show that this type of exported protein requires unfolding to move from the parasitophorous vacuole (PV) into the host cell. This indicates that soluble PNEPs, like PEXEL/HT proteins, are exported by translocation across the PV membrane (PVM), highlighting protein translocation in the parasite periphery as a general means in protein export of malaria parasites.

Proteomic and genetic analyses demonstrate that Plasmodium berghei blood stages export a large and diverse repertoire of proteins.

Malaria parasites actively remodel the infected red blood cell (irbc) by exporting proteins into the host cell cytoplasm. The human parasite Plasmodium falciparum exports particularly large numbers of proteins, including proteins that establish a vesicular network allowing the trafficking of proteins onto the surface of irbcs that are responsible for tissue sequestration. Like P. falciparum, the rodent parasite P. berghei ANKA sequesters via irbc interactions with the host receptor CD36. We have applied proteomic, genomic, and reverse-genetic approaches to identify P. berghei proteins potentially involved in the transport of proteins to the irbc surface. A comparative proteomics analysis of P. berghei non-sequestering and sequestering parasites was used to determine changes in the irbc membrane associated with sequestration. Subsequent tagging experiments identified 13 proteins (Plasmodium export element (PEXEL)-positive as well as PEXEL-negative) that are exported into the irbc cytoplasm and have distinct localization patterns: a dispersed and/or patchy distribution, a punctate vesicle-like pattern in the cytoplasm, or a distinct location at the irbc membrane. Members of the PEXEL-negative BIR and PEXEL-positive Pb-fam-3 show a dispersed localization in the irbc cytoplasm, but not at the irbc surface. Two of the identified exported proteins are transported to the irbc membrane and were named erythrocyte membrane associated proteins. EMAP1 is a member of the PEXEL-negative Pb-fam-1 family, and EMAP2 is a PEXEL-positive protein encoded by a single copy gene; neither protein plays a direct role in sequestration. Our observations clearly indicate that P. berghei traffics a diverse range of proteins to different cellular locations via mechanisms that are analogous to those employed by P. falciparum. This information can be exploited to generate transgenic humanized rodent P. berghei parasites expressing chimeric P. berghei/P. falciparum proteins on the surface of rodent irbc, thereby opening new avenues for in vivo screening adjunct therapies that block sequestration.

Multidimensional Design of Anticancer Peptides.

The computer-assisted design and optimization of peptides with selective cancer cell killing activity was achieved through merging the features of anticancer peptides, cell-penetrating peptides, and tumor-homing peptides. Machine-learning classifiers identified candidate peptides that possess the predicted properties. Starting from a template amino acid sequence, peptide cytotoxicity against a range of cancer cell lines was systematically optimized while minimizing the effects on primary human endothelial cells. The computer-generated sequences featured improved cancer-cell penetration, induced cancer-cell apoptosis, and were enabled a decrease in the cytotoxic concentration of co-administered chemotherapeutic agents in vitro. This study demonstrates the potential of multidimensional machine-learning methods for rapidly obtaining peptides with the desired cellular activities.

Piloting the membranolytic activities of peptides with a self-organizing map.

Antimicrobial peptides (AMPs) show remarkable selectivity toward lipid membranes and possess promising antibiotic potential. Their modes of action are diverse and not fully understood, and innovative peptide design strategies are needed to generate AMPs with improved properties. We present a de novo peptide design approach that resulted in new AMPs possessing low-nanomolar membranolytic activities. Thermal analysis revealed an entropy-driven mechanism of action. The study demonstrates sustained potential of advanced computational methods for designing peptides with the desired activity.

Scrutinizing MHC-I binding peptides and their limits of variation.

Designed peptides that bind to major histocompatibility protein I (MHC-I) allomorphs bear the promise of representing epitopes that stimulate a desired immune response. A rigorous bioinformatical exploration of sequence patterns hidden in peptides that bind to the mouse MHC-I allomorph H-2K(b) is presented. We exemplify and validate these motif findings by systematically dissecting the epitope SIINFEKL and analyzing the resulting fragments for their binding potential to H-2K(b) in a thermal denaturation assay. The results demonstrate that only fragments exclusively retaining the carboxy- or amino-terminus of the reference peptide exhibit significant binding potential, with the N-terminal pentapeptide SIINF as shortest ligand. This study demonstrates that sophisticated machine-learning algorithms excel at extracting fine-grained patterns from peptide sequence data and predicting MHC-I binding peptides, thereby considerably extending existing linear prediction models and providing a fresh view on the computer-based molecular design of future synthetic vaccines. The server for prediction is available at http://modlab-cadd.ethz.ch (SLiDER tool, MHC-I version 2012).

Prospective de novo drug design with deep interactome learning.

De novo drug design aims to generate molecules from scratch that possess specific chemical and pharmacological properties. We present a computational approach utilizing interactome-based deep learning for ligand- and structure-based generation of drug-like molecules. This method capitalizes on the unique strengths of both graph neural networks and chemical language models, offering an alternative to the need for application-specific reinforcement, transfer, or few-shot learning. It enables the "zero-shot" construction of compound libraries tailored to possess specific bioactivity, synthesizability, and structural novelty. In order to proactively evaluate the deep interactome learning framework for protein structure-based drug design, potential new ligands targeting the binding site of the human peroxisome proliferator-activated receptor (PPAR) subtype gamma are generated. The top-ranking designs are chemically synthesized and computationally, biophysically, and biochemically characterized. Potent PPAR partial agonists are identified, demonstrating favorable activity and the desired selectivity profiles for both nuclear receptors and off-target interactions. Crystal structure determination of the ligand-receptor complex confirms the anticipated binding mode. This successful outcome positively advocates interactome-based de novo design for application in bioorganic and medicinal chemistry, enabling the creation of innovative bioactive molecules.

Adaptive peptide design.

Computer algorithms help in the identification and optimization of peptides with desired structure and function. We provide an overview of the current focus of our research group in this field, highlighting innovative methods for peptide representation and de novo peptide generation. Our evolutionary molecular design cycle contains structure-activity relationship modeling by machine-learning methods, virtual peptide generation, activity prediction, peptide syntheses, as well as biophysical and biochemical activity determination. Such interplay between computer-assisted peptide generation and scoring with real laboratory experiments enables rapid feedback throughout the design cycle so that adaptive optimization can take place. Selected practical applications are reviewed including the design of new immunomodulatory MHC-I binding peptides and antimicrobial peptides.

Architecture, function and prediction of long signal peptides.

Protein targeting in eukaryotic cells is vital for cell survival and development. N-terminal signal peptides guide proteins to the membrane of the endoplasmic reticulum (ER) and initiate translocation into the ER lumen. Here, we review the status of signal peptide architecture and prediction with an emphasis on exceptionally long signal peptides, which often escape the notion of the currently available prediction methods. We benchmark publicly available prediction methods for their ability to correctly identify exceptionally long signal peptides. A set of 136 annotated eukaryotic signals served as reference data. The best prediction tool detected only 63%. A potential reason for the poor performance is the domain architecture of long signal peptides, whose structural peculiarities are insufficiently considered by current prediction algorithms. To overcome this limitation, we motivate a general domain view of long signal peptides, which becomes detectable when both the overall length and secondary structure of long signal peptides are taken into consideration. This concept provides a structural framework for identifying and understanding multiple targeting and post-targeting functions.

Adhesion, invasion, and agglutination mediated by two trimeric autotransporters in the human uropathogen Proteus mirabilis.

Fimbriae of the human uropathogen Proteus mirabilis are the only characterized surface proteins that contribute to its virulence by mediating adhesion and invasion of the uroepithelia. PMI2122 (AipA) and PMI2575 (TaaP) are annotated in the genome of strain HI4320 as trimeric autotransporters with "adhesin-like" and "agglutinating adhesin-like" properties, respectively. The C-terminal 62 amino acids (aa) in AipA and 76 aa in TaaP are homologous to the translocator domains of YadA from Yersinia enterocolitica and Hia from Haemophilus influenzae. Comparative protein modeling using the Hia three-dimensional structure as a template predicted that each of these domains would contain four antiparallel beta sheets and that they formed homotrimers. Recombinant AipA and TaaP were seen as ∼28 kDa and ∼78 kDa, respectively, in Escherichia coli, and each also formed high-molecular-weight homotrimers, thus supporting this model. E. coli synthesizing AipA or TaaP bound to extracellular matrix proteins with a 10- to 60-fold-higher level of affinity than the control strain. Inactivation of aipA in P. mirabilis strains significantly (P < 0.01) reduced the mutants' ability to adhere to or invade HEK293 cell monolayers, and the functions were restored upon complementation. A 51-aa-long invasin region in the AipA passenger domain was required for this function. E. coli expressing TaaP mediated autoagglutination, and a taaP mutant of P. mirabilis showed significantly (P < 0.05) more reduced aggregation than HI4320. Gly-247 in AipA and Gly-708 in TaaP were indispensable for trimerization and activity. AipA and TaaP individually offered advantages to P. mirabilis in a murine model. This is the first report characterizing trimeric autotransporters in P. mirabilis as afimbrial surface adhesins and autoagglutinins.

MHC I stabilizing potential of computer-designed octapeptides.

Experimental results are presented for 180 in silico designed octapeptide sequences and their stabilizing effects on the major histocompatibility class I molecule H-2K(b). Peptide sequence design was accomplished by a combination of an ant colony optimization algorithm with artificial neural network classifiers. Experimental tests yielded nine H-2K(b) stabilizing and 171 nonstabilizing peptides. 28 among the nonstabilizing octapeptides contain canonical motif residues known to be favorable for MHC I stabilization. For characterization of the area covered by stabilizing and non-stabilizing octapeptides in sequence space, we visualized the distribution of 100,603 octapeptides using a self-organizing map. The experimental results present evidence that the canonical sequence motives of the SYFPEITHI database on their own are insufficient for predicting MHC I protein stabilization.

An unusual ERAD-like complex is targeted to the apicoplast of Plasmodium falciparum.

Many apicomplexan parasites, including Plasmodium falciparum, harbor a so-called apicoplast, a complex plastid of red algal origin which was gained by a secondary endosymbiotic event. The exact molecular mechanisms directing the transport of nuclear-encoded proteins to the apicoplast of P. falciparum are not well understood. Recently, in silico analyses revealed a second copy of proteins homologous to components of the endoplasmic reticulum (ER)-associated protein degradation (ERAD) system in organisms with secondary plastids, including the malaria parasite P. falciparum. These proteins are predicted to be endowed with an apicoplast targeting signal and are suggested to play a role in the transport of nuclear-encoded proteins to the apicoplast. Here, we have studied components of this ERAD-derived putative preprotein translocon complex in malaria parasites. Using transfection technology coupled with fluorescence imaging techniques we can demonstrate that the N terminus of several ERAD-derived components targets green fluorescent protein to the apicoplast. Furthermore, we confirm that full-length PfsDer1-1 and PfsUba1 (homologues of yeast ERAD components) localize to the apicoplast, where PfsDer1-1 tightly associates with membranes. Conversely, PfhDer1-1 (a host-specific copy of the Der1-1 protein) localizes to the ER. Our data suggest that ERAD components have been "rewired" to provide a conduit for protein transport to the apicoplast. Our results are discussed in relation to the nature of the apicoplast protein transport machinery.