Wet Lab Validation Research Articles

A structure-based framework for selective inhibitor design and optimization

Structure-based drug design aims to create active compounds with favorable properties by analyzing target structures. Recently, deep generative models have facilitated structure-specific molecular generation. However, many methods are limited by inadequate pharmaceutical data, resulting in suboptimal molecular properties and unstable conformations. Additionally, these approaches often overlook binding pocket interactions and struggle with selective inhibitor design. To address these challenges, we developed a framework called Coarse-grained and Multi-dimensional Data-driven molecular generation (CMD-GEN). CMD-GEN bridges ligand-protein complexes with drug-like molecules by utilizing coarse-grained pharmacophore points sampled from diffusion model, enriching training data. Through a hierarchical architecture, it decomposes three-dimensional molecule generation within the pocket into pharmacophore point sampling, chemical structure generation, and conformation alignment, mitigating instability issues. CMD-GEN outperforms other methods in benchmark tests and controls drug-likeness effectively. Furthermore, CMD-GEN excels in cases across three synthetic lethal targets, and wet-lab validation with PARP1/2 inhibitors confirms its potential in selective inhibitor design.

Exploring the anti-diabetic potential of barnyard millet: insights from virtual screening, MD simulation and MM-PBSA

Barnyard millet (Echinochloa frumentacea) is a nutritionally superior grain and a rich source of dietary fiber, and protein. It helps in managing health and dietary issues such as malnutrition, diabetes, obesity, and celiac disease. Its low content of slowly digestible carbohydrates promotes a gradual release of glucose, helping to maintain stable blood glucose levels. The present study aims to identify and screen phytochemicals in the barnyard millet and explore its anti-diabetic activity through an in-silico study. Gas chromatography-mass spectrometry (GC-MS) analyses of the seed extract revealed the occurrence of 73 bioactive compounds that are known to possess a variety of pharmacological activities. Based on the virtual screening analysis, phytochemicals interacted with five different diabetic targets, with diosgenin demonstrating the lowest binding affinity across four receptors. Specifically, diosgenin showed a binding affinity of −9.2 kcal/mol with the Insulin receptor (PDB ID: 1IR3), −8.7 kcal/mol with Peroxisome proliferator-activated receptors (PDB ID: 3G9E), −7.5 kcal/mol with Tyrosine phosphatase 1-beta (2F70), and −6.5 kcal/mol with the Glucagon receptor (PDB ID: 5EE7). For Aldose reductase (PDB ID: 4XZH), Docosahexaenoic acid exhibited the lowest binding affinity of −9.9 kcal/mol. The dynamic behavior of 2F70-Diosgenin docked complexes throughout a 500 ns trajectory run was investigated further. The RMSD and RMSF analyses reveal that the complex remains structurally stable. The binding free energies were computed using the Molecular Mechanics/Poisson-Boltzmann Surface Area (MM/PBSA) methodology. The calculation results show that the predicted free energies of the complex are stable. These results suggest that the 2F70-Diosgenin complex is stable, highlighting its potential for further wet lab validation.

In-silico and in-vitro mutational analysis to investigate the structure and pathogenicity of POT1 missense variants

Abstract We performed an initial assessment of 24 missense POT1 variants using variant effect predictors (VEPs) followed by an extensive molecular dynamics (MD) simulations of 14 selected variants: N-terminal - I115T, Y223C, D224N, M251V, H266Q, S270N, C-terminal - A364E, P371T, T416A, S421L, A532P, L538V, P601L, Q623H for the unbound state and bound state with telomeric ssDNA (N-terminal) or TPP-PBD (C-terminal). Wet-lab validations using electrophoretic mobility shift assay (EMSA), telomere dysfunction-induced foci (TIF) formation analysis, and co-immunoprecipitation were done to further explore the functional impacts. Initial assessment included VEPs, FoldX, and I-Mutant2.0 followed by MD analysis which revealed structural instability and reduction in associated binding efficiency in D224N, H266Q, S270N. EMSA showed remarkable reduction of POT1-telomere oligo binding in D224N, H266Q, and S270N. Through TIF formation analysis, H266Q and P371T exhibited the highest number of co-localized foci compared to other mutant and WT cell lines. Also, POT1-(FLAG)-TPP1 co-immunoprecipitation suggested a reduction in POT1-TPP1 binding in all C-terminal mutants, most notably in P601L and Q623H. Our study demonstrated a comprehensive workflow to characterize missense POT1 variants of uncertain significance and hence proposing this workflow to be considered in the American College of Medical Genetics and Genomics (ACMG) criteria for missense POT1 variant classification.

Role of Artificial Intelligence in Drug Discovery to Revolutionize the Pharmaceutical Industry: Resources, Methods and Applications.

Traditional drug discovery methods such as wet-lab testing, validations, and synthetic techniques are time-consuming and expensive. Artificial Intelligence (AI) approaches have progressed to the point where they can have a significant impact on the drug discovery process. Using massive volumes of open data, artificial intelligence methods are revolutionizing the pharmaceutical industry. In the last few decades, many AI-based models have been developed and implemented in many areas of the drug development process. These models have been used as a supplement to conventional research to uncover superior pharmaceuticals expeditiously. AI's involvement in the pharmaceutical industry was used mostly for reverse engineering of existing patents and the invention of new synthesis pathways. Drug research and development to repurposing and productivity benefits in the pharmaceutical business through clinical trials. AI is studied in this article for its numerous potential uses. We have discussed how AI can be put to use in the pharmaceutical sector, specifically for predicting a drug's toxicity, bioactivity, and physicochemical characteristics, among other things. In this review article, we have discussed its application to a variety of problems, including de novo drug discovery, target structure prediction, interaction prediction, and binding affinity prediction. AI for predicting drug interactions and nanomedicines were also considered.

Alternative splicing of CADM1 is associated with endothelial progenitor cell dysfunction in preeclampsia.

Preeclampsia is a pregnancy specific hypertensive disorder and is associated with an increased postpartum risk of cardiovascular morbidity for both women and their offspring. Previous studies have indicated that cord blood endothelial colony forming cells (ECFCs) are dysfunctional in preeclampsia. The specific mechanisms are not yet fully understood but dysregulation of alternative splicing has been proposed as one of the pathogenic pathways. In order to identify specific targets of alternative splicing in fetal ECFCs, we performed transcriptome-wide differential splicing analyses between cord blood ECFCs from preeclamptic (n=16) and normal pregnancies (n=13). Selected splicing events were validated using fragment length analysis and Sanger sequencing. In silico transcriptome-wide differential splicing analysis identified a significantly increased abundance of the CADM1 isoform ENST00000542447 in the preeclamptic cohort (p=0.002), which was confirmed by wet-lab validation. The deleted exon 8 harbours glycosylation sites known to mediate cell-cell adhesion. To investigate the functional impact of alternative splice variants, we induced an in vitro splice switch using antisense morpholino treatment and followed cellular effects using migration and angiogenesis assays in ECFCs from six normal pregnancies. The CADM1 exon 8 skipping converted the normal ECFCs to a preeclampsia-like state characterised by a decreased migration ability (panova=0.005) and decreased tubule length (panova=0.02). We propose aberrant splicing of CADM1 and the resulting changes in the adherence properties of ECFCs as a potential contributor to cardiovascular sequelae in the offspring of preeclamptic pregnancies.

Discovery of anticancer peptides from natural and generated sequences using deep learning.

Discovery of anticancer peptides from natural and generated sequences using deep learning.

Maternal nutrition and offspring lung health: sex-specific pathway modulation in fibrosis, metabolism, and immunity.

Maternal nutrition profoundly influences offspring health, impacting both prenatal and early postnatal development. Previous studies have demonstrated that maternal dietary habits can affect key developmental pathways in the offsprings, including those related to lung function and disease susceptibility. However, the sex-specific impact of a maternal high-salt diet (HSD) on offspring lung injury remains poorly understood. This study aimed to investigate the sex-specific effects of maternal HSD on lung injury in mouse offsprings, focusing on pathways related to fibrosis, metabolism, immunity, and apoptosis. Pregnant C57BL/6J mice were subjected to either normal or HSD conditions during gestation. Lung tissues from the male and female offsprings were analyzed using high-throughput RNA sequencing and bioinformatics tools to examine transcriptomic changes. Wet-lab validation, including Masson trichrome staining, immunofluorescence for α-SMA, and qRT-PCR for fibrotic markers (α-SMA, collagen I, Fn1, and TGF-β), was conducted to confirm fibrosis and other injury markers. Lung structure and weight were also evaluated to assess physical alterations due to maternal diet. Maternal HSD significantly altered lung transcriptomes in a sex-specific manner. Male offsprings showed increased susceptibility to fibrosis, as confirmed by histological and molecular analyses, including elevated expression of α-SMA, collagen I, Fn1, and TGF-β. In contrast, female offsprings exhibited distinct changes in metabolic and immune pathways. These findings highlight the differential regulation of pulmonary injury mechanisms between male and female offsprings exposed to HSD. Maternal HSD induces sex-specific lung injury in offsprings by disrupting critical pathways involved in fibrosis, metabolism, immunity, and apoptosis. The combination of transcriptomic and orthogonal data underscores the need for balanced maternal nutrition during pregnancy to promote long-term respiratory health in offsprings. These results provide new insights into the sex-specific vulnerabilities to lung disease arising from maternal diet.

Immunoinformatic evaluation for the development of a potent multi-epitope vaccine against bacterial vaginosis caused by Gardnerella vaginalis.

Bacterial vaginosis (BV) is the most common vaginal dysbiosis in fertile women, which is associated with side effects including the risk of premature birth. Gardnerella vaginalis (G. vaginalis) is a facultative anaerobic bacillus known as the main pathogen responsible for BV. In this study, using bioinformatics and immunoinformatics methods, a multi-epitope vaccine with optimal population coverage against BV caused by G. vaginalis was designed. Amino acid sequences of two important virulence factors (Vaginolysin and Sialidase) of G. vaginalis were retrieved from NCBI and UniProt databases. At first, three online servers ABCpred, BCPREDS and LBtope were used to predict linear B-cell epitopes (BCEs) and IEDB server was used for T cells. Then the antigenicity, toxicity, allergenicity were evaluated using bioinformatics tools. After modeling the three-dimensional (3D) structure of the vaccine by Robetta Server, molecular docking and molecular dynamics were performed. Finally, immune simulation and in silico cloning were considered effective for the design of vaccine production strategy. In total, six epitopes of BCEs, eight epitopes from CD4+ and seven epitopes from CD8+ were selected. The designed multi-epitope vaccine was non-allergenic and non-toxic and showed high levels of antigenicity and immunogenicity. After the 3D structure was predicted, it was refined and validated, which resulted in an optimized model with a Z-score of -7.4. Molecular docking and molecular dynamics simulation of the designed vaccine revealed stable and strong binding interactions. Finally, the results of vaccine immunity simulation showed a significant increase in immunoglobulins, higher levels of IFN-γ and IL-2. According to the findings, the candidate multi-epitope vaccine has stable structural features. It also has the potential to stimulate long-term immunity in the host, but wet-lab validation is needed to justify it.

Design, validation, and implementation of multiplex digital PCR assays for simultaneous quantification of multiple targets.

Quantitative polymerase chain reaction (qPCR) and digital PCR (dPCR) are applied for quantifying molecular targets in disease diagnostics, pathogen detection, and ecological monitoring. Uptake of dPCR is increasing due to its higher quantification accuracy relative to qPCR, which stems from its independence from standard curves and its increased resistance to PCR inhibitors. Throughput can be increased through multiplexing, which allows simultaneous quantification of multiple targets. However, multiplexing with dPCR faces unique challenges relative to qPCR. Here, we describe the three-phase development process of non-competing multiplex dPCR assays using target-specific fluorescently labeled hydrolysis probes. We highlight common challenges encountered, along with recommended solutions. Phase 1: In silico assay design; target-specific primers and probes are selected or designed, potential issues with primer and probe interactions are identified, and fluorophores and quenchers are chosen based on dPCR instrumentation. Phase 2: Wet-lab validation; assays are benchmarked using positive controls. Insufficient performance leads to assay redesign, as needed. Phase 3: Assay implementation; assay specificity and sensitivity are validated on relevant sample matrices. Finally, we provide recommendations on the future design and standardization of multiplexed dPCR assays, highlighting the need for better in silico predictions of assay performance, standardizing positive controls, and automating partition classification systems.

Unlocking the anti-cancer potential of phytocompounds from essential oils extracted from Cymbopogon martini, Melaleuca alternifolia, Rosmarinus officinalis and Boswellia serrate: in-silico docking and wet lab validation

Cymbopogon martini, Melaleuca alternifolia, Rosmarinus officinalis and Boswellia serrata are medicinal plants that include ingredients that have therapeutic and medicinal phytocompounds endowed with innumerable pharmacological and biological activities. These plants are widely known for producing herbal essential oils (EOs) that are used as traditional and complementary medicine to cure various ailments. This study aimed in-silico docking of major phytoconstituents against cancer associated cancer-related human epidermal growth factor receptor (HER2) and mammalian target of rapamycin (mTORC2) proteins as relevant targets for cancer medication development followed by wet lab validation using in vitro approach. The hydro-distillation process was used to extract the EOs from plants. Subsequently, the most dominant compounds along with some minor compounds from all EOs were docked using computational study. Subsequently, an in-vitro confirmation study was conducted to ascertain each EO's potential for anticancer effects. The (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) MTT test measured the essential oil's anti-cancer effect on HeLa cell line. A study on EO’s antioxidant properties was also conducted. The pharmacokinetic characteristics were also established. Findings from the in-silico investigation indicated a strong docking of phytochemicals against HER2 and mTORC2 cancer cell targets. The ADMET prediction tool was utilized to estimate the in-silico pharmacokinetic and drug-like characteristics of the key compounds. The observed findings shown that all phytocompounds have drug like properties. Finally, during wet lab validation, EOs were shown to exhibit robust cytotoxicity against the HeLa cell line after 24 h incubation. Moreover, EOs possess potent antioxidant properties observed by 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,2′-azino-bis-(3-ethylbenzothiazoline-6-sulfonic) acid (ABTS), Hydroxyl Radical scavenging and Iron Chelating Assays. These foremost findings may help in the development of novel anti-proliferative treatment drugs. Therefore, more research is necessary to establish the in vivo anticancer properties of this species of Cymbopogon martinii, Melaleuca alternifolia, Rosmarinus officinalis, and Boswellia serrata in cancer cells as well as to investigate the molecular processes behind their anticancer activities.

ClickGen: Directed exploration of synthesizable chemical space via modular reactions and reinforcement learning

Despite the significant potential of generative models, low synthesizability of many generated molecules limits their real-world applications. In response to this issue, we develop ClickGen, a deep learning model that utilizes modular reactions like click chemistry to assemble molecules and incorporates reinforcement learning along with inpainting technique to ensure that the proposed molecules display high diversity, novelty and strong binding tendency. ClickGen demonstrates superior performance over the other reaction-based generative models in terms of novelty, synthesizability, and docking conformation similarity for existing binders targeting the three proteins. We then proceeded to conduct wet-lab validation on the ClickGen’s proposed molecules for poly adenosine diphosphate-ribose polymerase 1. Due to the guaranteed high synthesizability and model-generated synthetic routes for reference, we successfully produced and tested the bioactivity of these novel compounds in just 20 days, much faster than typically expected time frame when handling sufficiently novel molecules. In bioactivity assays, two lead compounds demonstrated superior anti-proliferative efficacy against cancer cell lines, low toxicity, and nanomolar-level inhibitory activity to PARP1. We demonstrate that ClickGen and related models may represent a new paradigm in molecular generation, bringing AI-driven, automated experimentation and closed-loop molecular design closer to realization.

Machine learning for potential anti-cancer discovery from black sea cucumbers

<p>Despite being an abundant marine organism in Indonesia, black sea cucumbers <em>(Holothuria atra)</em> is still underutilised due to its slightly bitter taste. This study aims to identify potential anti-cancer compounds from black sea cucumbers using machine learning (ML) to perform drug discovery. ML models were used to predict interactions between compounds from the organism with cancer-related proteins. Following prediction, all compounds were computationally validated through molecular docking. The validated compounds were then screened using absorption, distribution, metabolism, excretion, and toxicity (ADMET) Lab 2.0 to assess their druglike properties. The results showed that ML predicted seven out of 86 compounds were interacted with cancer-related proteins. Computational validation from the results showed that four out of seven compounds demonstrated stable interaction with proteins where only one compound meet the criteria of drug-like compound. The framework of ML and computational validation highlighted in this study shows a great promise in the future of drug discovery specifically for marine organisms. Since computational method only works in prediction realms, wet lab validation and clinical trials are imperative before the drug candidate can be produced as actual anti-cancer drug.</p>

Targeting Monkeypox Virus Methyltransferase: Virtual Screening of Natural Compounds from Middle-Eastern Medicinal Plants.

Monkeypox is an infectious disease resulting from the monkeypox virus, and its fatality rate varies depending on the virus clade and the location of the outbreak. In monkeypox virus, methyltransferase (MTase) plays a crucial role in modifying the cap structure of viral mRNA. This alteration assists the virus in evading the host's immune system, enhances viral protein synthesis, and ultimately enables successful infection and replication within host cells. Given the significance of MTase in viral infection and spread within the host, our study aimed to identify a natural inhibitor for this enzyme using docking and molecular dynamic (MD) simulations. We collected a total of 12,971 natural compounds from 200 medicinal plants in the Middle East. After eliminating duplicate compounds, we had 5,749 unique ligand conformers, which we then subjected to high-throughput virtual screening against MTase. The most promising hits were further evaluated using the extra-precision (XP) tool. The affinity of these hits was also assessed by Prime-Molecular Mechanics/Generalized Born Surface Area (MMGBSA) tool. The analysis revealed that two standard controls (sinefungin and TO1119) and two Middle-Eastern compounds (folic acid and 1,2,4,6-tetragalloylglucose) exhibited the best XP docking scores. According to Prime MMGBSA calculations, the Middle-Eastern compounds showed higher affinities, with values of -60.61 kcal/mol for 1,2,4,6-tetragalloylglucose and -51.87 kcal/mol for folic acid, surpassing the controls (TO1119 at -35.71 kcal/mol and sinefungin at -31.51 kcal/mol). In the majority of Molecular dynamic (MD) simulations, folic acid exhibited demonstrated greater stability than sinefungin. Further investigation revealed that folic acid occupied a critical position in the active site of MTase, which reduced its interaction with the mRNA substrate. Based on these findings, it can be concluded that folic acid is a highly promising natural compound for potential use in the cost-effective treatment of monkeypox virus. The identification of folic acid as a potential antiviral agent highlights the importance of nature in providing new therapeutic uses that have significant implications for global health, particularly in regions where monkeypox viral outbreaks are prevalent. However, it is essential to note that further wet-lab validations are necessary to confirm its efficacy for treatment in a medical context.

Secondary Analysis of Human Bulk RNA-Seq Dataset Suggests Potential Mechanisms for Letrozole Resistance in Estrogen-Positive (ER+) Breast Cancer.

Estrogen receptor-positive (ER+) breast cancer is common among postmenopausal women and is frequently treated with Letrozole, which inhibits aromatase from synthesizing estrogen from androgens. Decreased estrogen slows the growth of tumors and can be an effective treatment. The increase in Letrozole resistance poses a unique problem for patients. To better understand the underlying molecular mechanism(s) of Letrozole resistance, we reanalyzed transcriptomic data by comparing individuals who responded to Letrozole therapy (responders) to those who were resistant to treatment (non-responders). We identified SOX11 and S100A9 as two significant differentially expressed genes (DEGs) between these patient cohorts, with "PLK1 signaling events" being the most significant signaling pathway. We also identified PRDX4 and E2F8 gene products as being the top mechanistic transcriptional markers for ER+ treatment resistance. Many of the significant DEGs that we identified play a known role in ER+ breast cancer or other types of cancer, which partially validate our results. Several of the gene products we identified are novel in the context of ER+ breast cancer. Many of the genes that we identified warrant further research to elucidate the more specific molecular mechanisms of Letrozole resistance in this patient population and could potentially be used as prognostic markers with further wet lab validation. We anticipate that these findings could contribute to improved detection and therapeutic outcomes in aromatase-resistant ER+ breast cancer patients.

LncRNA HULC augments high glucose-associated pancreatic cancer progression and drug resistance by enhancing YAP activity and autophagy.

LncRNA HULC augments high glucose-associated pancreatic cancer progression and drug resistance by enhancing YAP activity and autophagy.

Computational investigation of Y. aloifolia variegate as anti-Human Immunodeficiency Virus (HIV) targeting HIV-1 protease: A multiscale in-silico exploration

Computational investigation of Y. aloifolia variegate as anti-Human Immunodeficiency Virus (HIV) targeting HIV-1 protease: A multiscale in-silico exploration

A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage and its implications.

The manufacturability assessment and optimization of bispecific antibodies (bsAbs) during the discovery stage are crucial for the success of the drug development process, impacting the speed and cost of advancing such therapeutics to the Investigational New Drug (IND) stage and ultimately to the market. The complexity of bsAbs creates challenges in employing effective evaluation methods to detect developability risks in early discovery stage, and poses difficulties in identifying the root causes and implementing subsequent engineering solutions. This study presents a case of engineering a bsAb that displayed a normal solution appearance during the discovery phase but underwent significant precipitation when subjected to agitation stress during 15L Chemistry, Manufacturing, and Control (CMC) production Leveraging analytical tools, structural analysis, in silico prediction, and wet-lab validations, the key molecular origins responsible for the observed precipitation were identified and addressed. Sequence engineering to reduce protein surface hydrophobicity and enhance conformational stability proved effective in resolving agitation-induced aggregation. The refined bsAb sequences enabled successful mass production in CMC department. The findings of this case study contribute to the understanding of the fundamental mechanism of agitation-induced aggregation and offer a potential protein engineering procedure for addressing similar issues in bsAb. Furthermore, this case study emphasizes the significance of a close partnership between Discovery and CMC teams. Integrating CMC's rigorous evaluation methods with Discovery's engineering capability can facilitate a streamlined development process for bsAb molecules.

Enhancing Gene Co-Expression Network Inference for the Malaria Parasite Plasmodium falciparum.

Malaria results in more than 550,000 deaths each year due to drug resistance in the most lethal Plasmodium (P.) species P. falciparum. A full P. falciparum genome was published in 2002, yet 44.6% of its genes have unknown functions. Improving the functional annotation of genes is important for identifying drug targets and understanding the evolution of drug resistance. Genes function by interacting with one another. So, analyzing gene co-expression networks can enhance functional annotations and prioritize genes for wet lab validation. Earlier efforts to build gene co-expression networks in P. falciparum have been limited to a single network inference method or gaining biological understanding for only a single gene and its interacting partners. Here, we explore multiple inference methods and aim to systematically predict functional annotations for all P. falciparum genes. We evaluate each inferred network based on how well it predicts existing gene-Gene Ontology (GO) term annotations using network clustering and leave-one-out crossvalidation. We assess overlaps of the different networks' edges (gene co-expression relationships), as well as predicted functional knowledge. The networks' edges are overall complementary: 47-85% of all edges are unique to each network. In terms of the accuracy of predicting gene functional annotations, all networks yielded relatively high precision (as high as 87% for the network inferred using mutual information), but the highest recall reached was below 15%. All networks having low recall means that none of them capture a large amount of all existing gene-GO term annotations. In fact, their annotation predictions are highly complementary, with the largest pairwise overlap of only 27%. We provide ranked lists of inferred gene-gene interactions and predicted gene-GO term annotations for future use and wet lab validation by the malaria community. The different networks seem to capture different aspects of the P. falciparum biology in terms of both inferred interactions and predicted gene functional annotations. Thus, relying on a single network inference method should be avoided when possible. Attached.

Integrated microRNA and proteome analysis of cancer datasets with MoPC.

MicroRNAs (miRNAs) are small molecules that play an essential role in regulating gene expression by post-transcriptional gene silencing. Their study is crucial in revealing the fundamental processes underlying pathologies and, in particular, cancer. To date, most studies on miRNA regulation consider the effect of specific miRNAs on specific target mRNAs, providing wet-lab validation. However, few tools have been developed to explain the miRNA-mediated regulation at the protein level. In this paper, the MoPC computational tool is presented, that relies on the partial correlation between mRNAs and proteins conditioned on the miRNA expression to predict miRNA-target interactions in multi-omic datasets. MoPC returns the list of significant miRNA-target interactions and plot the significant correlations on the heatmap in which the miRNAs and targets are ordered by the chromosomal location. The software was applied on three TCGA/CPTAC datasets (breast, glioblastoma, and lung cancer), returning enriched results in three independent targets databases.

Extensive antibody search with whole spectrum black-box optimization

In designing functional biological sequences with machine learning, the activity predictor tends to be inaccurate due to shortage of data. Top ranked sequences are thus unlikely to contain effective ones. This paper proposes to take prediction stability into account to provide domain experts with a reasonable list of sequences to choose from. In our approach, multiple prediction models are trained by subsampling the training set and the multi-objective optimization problem, where one objective is the average activity and the other is the standard deviation, is solved. The Pareto front represents a list of sequences with the whole spectrum of activity and stability. Using this method, we designed VHH (Variable domain of Heavy chain of Heavy chain) antibodies based on the dataset obtained from deep mutational screening. To solve multi-objective optimization, we employed our sequence design software MOQA that uses quantum annealing. By applying several selection criteria to 19,778 designed sequences, five sequences were selected for wet-lab validation. One sequence, 16 mutations away from the closest training sequence, was successfully expressed and found to possess desired binding specificity. Our whole spectrum approach provides a balanced way of dealing with the prediction uncertainty, and can possibly be applied to extensive search of functional sequences.