Human Protein Sequences Research Articles

Mapping the Multiscale Proteomic Organization of Cellular and Disease Phenotypes.

While the primary sequences of human proteins have been cataloged for over a decade, determining how these are organized into a dynamic collection of multiprotein assemblies, with structures and functions spanning biological scales, is an ongoing venture. Systematic and data-driven analyses of these higher-order structures are emerging, facilitating the discovery and understanding of cellular phenotypes. At present, knowledge of protein localization and function has been primarily derived from manual annotation and curation in resources such as the Gene Ontology, which are biased toward richly annotated genes in the literature. Here, we envision a future powered by data-driven mapping of protein assemblies. These maps can capture and decode cellular functions through the integration of protein expression, localization, and interaction data across length scales and timescales. In this review, we focus on progress toward constructing integrated cell maps that accelerate the life sciences and translational research.

Independent expansion, selection, and hypervariability of the TBC1D3 gene family in humans.

TBC1D3 is a primate-specific gene family that has expanded in the human lineage and has been implicated in neuronal progenitor proliferation and expansion of the frontal cortex. The gene family and its expression have been challenging to investigate because it is embedded in high-identity and highly variable segmental duplications. We sequenced and assembled the gene family using long-read sequencing data from 34 humans and 11 nonhuman primate species. Our analysis shows that this particular gene family has independently duplicated in at least five primate lineages, and the duplicated loci are enriched at sites of large-scale chromosomal rearrangements on Chromosome 17. We find that all human copy-number variation maps to two distinct clusters located at Chromosome 17q12 and that humans are highly structurally variable at this locus, differing by as many as 20 copies and ∼1 Mbp in length depending on haplotypes. We also show evidence of positive selection, as well as a significant change in the predicted human TBC1D3 protein sequence. Last, we find that, despite multiple duplications, human TBC1D3 expression is limited to a subset of copies and, most notably, from a single paralog group: TBC1D3-CDKL These observations may help explain why a gene potentially important in cortical development can be so variable in the human population.

Abstract B009: Identifying of a novel diagnostic markers for pancreatic neuroendocrine tumors by proteomics with patient blood

Abstract Neuroendocrine tumors are tumors originating from endocrine cells in various organs, and present very heterogeneous characteristics, and the incidence rate is gradually increasing. Among them, neuroendocrine tumors originating from pancreas are divided into functional tumors and non-functional tumors according to the presence hormone secretion function. Since non-functional tumors are often asymptomatic, early diagnosis is not easy, and differential diagnosis from pancreas ductal adenocarcinoma (PDAC) is difficult, making it difficult to determine the direction of treatment. Therefore, we aimed to find protein biomarkers for the diagnosis and differential diagnosis of pancreatic neuroendocrine tumors (PNET) through proteomic analysis of human blood. First, a total of 60 human plasma samples were obtained from 17 PNETs, 23 PDACs, and 20 Healthy donors (HD), and liquid chromatography-mass spectrometry (LC-MS) proteomics analysis was performed. For 60 cases of raw data obtained through LC-MS analysis, identification and label-free quantification were performed through the Human swissprot protein sequence database. Through pre-processing and normalization, an average of 549 protein lists for each sample were obtained. And through quantitative evaluation, the proteomes that were increased or decreased for each disease were analyzed and classified, and ACTR3, CD163, and LECT2, which were increased in PNET, were selected as diagnostic biomarker candidates. The diagnostic AUC for distinguishing PNET from HD was 0.706 for ACTR3, 0.971 for CD163, and 0.541 for LECT2. The diagnostic AUC to distinguish PNET from PDAC was 0.821 for ACTR3, 0.870 for CD163, and 0.783 for LECT2. We selected three candidates, ACTR3, CD163, and LECT2, as blood biomarkers for the accurate diagnosis of PNET. And through the combination of markers, it is expected that the sensitivity and accuracy of the diagnosis can be further increased. In the future, we will study the function as a marker that can simultaneously predict diagnosis and treatment response by identifying the molecular biological mechanism for the generation and progression of PNET through these three markers. Citation Format: Eun-Young Koh, Ji Hye Jeong, Hee-Sung Ahn, Kyunggon Kim, Eunsung Jun. Identifying of a novel diagnostic markers for pancreatic neuroendocrine tumors by proteomics with patient blood [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Pancreatic Cancer; 2023 Sep 27-30; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(2 Suppl):Abstract nr B009.

The distal-proximal relationships among the human moonlighting proteins: Evolutionary hotspots and Darwinian checkpoints

Moonlighting proteins, known for their ability to perform multiple, often unrelated functions within a single polypeptide chain, challenge the traditional “one gene, one protein, one function” paradigm. As organisms evolved, their genomes remained relatively stable in size, but the introduction of post-translational modifications and sub-strategies like protein promiscuity and intrinsic disorder enabled multifunctionality. Enzymes, in particular, exemplify this phenomenon, engaging in unrelated processes alongside their primary catalytic roles. This study employs a systematic, quantitative informatics approach to shed light on human moonlighting protein sequences. Phylogenetic analyses of human moonlighting proteins are presented, elucidating the distal-proximal relationships among these proteins based on sequence-derived quantitative features. The findings unveil the captivating world of human moonlighting proteins, urging further investigations in the emerging field of moonlighting proteomics, with the potential for significant contributions to our understanding of multifunctional proteins and their roles in diverse cellular processes and diseases.

Occurrences of allergenicity to banana pathogenesis-related-10 (PR10) protein variants.

Pathogenesis-related-10 (PR10) proteins play significant roles in plant defence against biotic and abiotic stresses. Recently, two banana PR10 proteins (MaPR10-BeB5 and MaPR10-GNA5) were characterised and shown to exhibit antifungal properties against Aspergillus fumigatus in vitro. In rice, transgenic overexpression of PR10 proteins conferred resistance to pathogen infection and drought tolerance without affecting productivity, highlighting their potential for agricultural applications. However, PR10 proteins also include the Bet v 1-like family of allergens implicated in pollen food allergy syndromes, raising concerns about potential adverse effects on human health. In this study, we evaluated the allergenic potential of the recently isolated banana PR10 proteins. We first predicted the presence of IgE epitopes of the Bet v 1 allergen family in the deduced PR10 peptide sequences in silico. We then predicted the structures of four human IgE scFv protein sequences and three plant PR10 protein sequences. Based on the quality of the predicted structures, one IgE scFv protein structure was selected for docking with the three plant PR10 proteins. We confirmed the docking results with immunoblot analysis performed using recombinant MaPR10-BeB5 and MaPR10-GNA5 proteins against the sera of banana-allergic patients. Our experimental results substantiated the notion that both protein variants are potentially allergenic since these proteins were recognised by 26.6% of banana-allergic patients with broad PR10 protein recognition. We caution that the allergenic potential of MaPR10 proteins should be carefully considered before implementing transgenic overexpression strategies to improve crops, with a suggestion to limit their expression to non-edible plant tissues.

Unique Patterns in Amino Acid Sequences of Aging-Related Proteins.

Aging has strong genetic components and the list of genes that may regulate the aging process is collected in the GenAge database. There may be characteristic patterns in the amino acid sequences of aging-related proteins that distinguish them from other proteins and this information will lead to a better understanding of the aging process. To test this hypothesis, human protein sequences are extracted from the UniProt database and the relative frequency of every amino acid residue in aging-related proteins and the remaining proteins is calculated. The main observation is that the mean relative frequency of aspartic acid (D) is consistently higher, while the mean relative frequencies of tryptophan (W) and leucine (L) are consistently lower in aging-related proteins compared to the non-aging-related proteins for the human and four examined model organisms. It is also observed that the mean relative frequency of aspartic acid is higher, while the mean relative frequency of tryptophan is lower in pro-longevity proteins compared to anti-longevity proteins in model organisms. Finally, it is found that aging-related proteins tend to be longer than non-aging-related proteins. It is hoped that this analysis initiates further computational and experimental research to explore the underlying mechanisms of these findings.

Conservation of Glutathione Transferase mRNA and Protein Sequences Similar to Human and Horse Alpha Class GST A3-3 across Dog, Goat, and Opossum Species.

The glutathione transferase A3-3 (GST A3-3) homodimeric enzyme is the most efficient enzyme that catalyzes isomerization of the precursors of testosterone, estradiol, and progesterone in the gonads of humans and horses. However, the presence of GST A3-3 orthologs with equally high ketosteroid isomerase activity has not been verified in other mammalian species, even though pig and cattle homologs have been cloned and studied. Identifying GSTA3 genes is a challenge because of multiple GSTA gene duplications (e.g., 12 in the human genome); consequently, the GSTA3 gene is not annotated in most genomes. To improve our understanding of GSTA3 gene products and their functions across diverse mammalian species, we cloned homologs of the horse and human GSTA3 mRNAs from the testes of a dog, goat, and gray short-tailed opossum, the genomes of which all currently lack GSTA3 gene annotations. The resultant novel GSTA3 mRNA and inferred protein sequences had a high level of conservation with human GSTA3 mRNA and protein sequences (≥70% and ≥64% identities, respectively). Sequence conservation was also apparent for the 12 residues of the "H-site" in the 222 amino acid GSTA3 protein that is known to interact with the steroid substrates. Modeling predicted that the dog GSTA3-3 may be a more active ketosteroid isomerase than the corresponding goat or opossum enzymes. However, expression of the GSTA3 gene was higher in liver than in other dog tissue. Our results improve understanding of the active sites of mammalian GST A3-3 enzymes, inhibitors of which might be useful for reducing steroidogenesis for medical purposes, such as fertility control or treatment of steroid-dependent diseases.

A normalized differential sequence feature encoding method based on amino acid sequences.

Protein interactions are the foundation of all metabolic activities of cells, such as apoptosis, the immune response, and metabolic pathways. In order to optimize the performance of protein interaction prediction, a coding method based on normalized difference sequence characteristics (NDSF) of amino acid sequences is proposed. By using the positional relationships between amino acids in the sequences and the correlation characteristics between sequence pairs, NDSF is jointly encoded. Using principal component analysis (PCA) and local linear embedding (LLE) dimensionality reduction methods, the coded 174-dimensional human protein sequence vector is extracted using sequence features. This study compares the classification performance of four ensemble learning methods (AdaBoost, Extra trees, LightGBM, XGBoost) applied to PCA and LLE features. Cross-validation and grid search methods are used to find the best combination of parameters. The results show that the accuracy of NDSF is generally higher than that of the sequence matrix-based coding method (MOS) coding method, and the loss and coding time can be greatly reduced. The bar chart of feature extraction shows that the classification accuracy is significantly higher when using the linear dimensionality reduction method, PCA, compared to the nonlinear dimensionality reduction method, LLE. After classification with XGBoost, the model accuracy reaches 99.2%, which provides the best performance among all models. This study suggests that NDSF combined with PCA and XGBoost may be an effective strategy for classifying different human protein interactions.

Presence of PD-1 similarity genes in monocytes may promote the development of type 1 diabetes mellitus and poor prognosis of pancreatic cancer

IntroductionTo identify proteins and corresponding genes that share sequential and structural similarity with programmed cell death protein-1 (PD-1) in patients with type 1 diabetes mellitus (T1DM) via bioinformatics analysis.Research design...

Abstract 1885: Selection of an optimal anti-PD-L1 single domain antibody format for the vectorization into oncolytic vaccinia virus and the generation of bispecific immunomodulators

Abstract Background: Single domain antibodies (sdAbs) isolated after immunization of camelids are particularly attractive formats for their high modularity and small size allowing a better diffusion into tumors. However, the short in vivo half-life of sdAbs, related to the lack of a Fc domain, limits their clinical application. By replicating specifically into tumor cells, the oncolytic vaccinia virus (VACV) is an optimal vector to deliver and maintain high intra-tumoral concentrations of therapeutic sdAb. Moreover, sdAb targeting immunological targets, such as PD-L1, may synergize the anti-tumoral activity of VACV. Randox and Transgene report the selection and characterization of a sdAb targeting the human PD-L1 and the design of optimal formats, including bispecific anti-PD-L1-CD40 ligand, for vectorization into VACV. Methods: Alpacas were immunized with human PD-L1 protein and sdAb coding sequences were isolated by PCR. Anti-PD-L1 sdAb binders were selected by phage display and sdAb blockers of PD-L1/PD-1 interaction were identified by ELISA. The ability of the selected sdAb to disrupt the PD-L1/PD-1 interaction was verified on transformed and primary cells. To fine-tune an optimal anti-PD-L1, several sdAb formats were designed and vectorized into VACV. The sdAb format exhibiting the best PD-L1/PD-1 blocking activity was selected by the screening of culture supernatants of several VACV-sdAb infected tumor cells. Finally, anti-PD-L1 sdAb-CD40L fusions were designed to generate a strong CD40 agonist active only in a PD-L1 positive environment. Results: SdAb clone 1A1 exhibited the best PD-L1/PD-1 blocking activity which remained unchanged after extensive humanization (latterly becoming named GS542). GS542 was vectorized in VACV as monomeric single chain homodimer, and fused to Fc domain or antibody hinge domain to foster dimerization together with full length IgG1 avelumab as anti-PD-L1 benchmark. All constructs were expressed by infected tumor cells. The single chain homodimer displayed the best PD-L1/PD-1 blocking activity, superior to that of avelumab. Moreover, GS542-CD40L fusions were designed to take advantage of the natural trimerization of CD40L to increase the avidity for PD-L1 while clustering CD40L at the surface of PD-L1+ cells to trans-activate the CD40 pathway. Evaluation of these GS542-CD40L fusions showed strong CD40 agonist activity depending on the presence of PD-L1+ cells making these constructs safer CD40 agonists. Conclusion: An anti-PD-L1 sdAb with a strong blocking activity was selected, humanized and evaluated under different VACV-vectorized formats. The single chain homodimeric sdAb expressed by VACV was identified as the best PD-L1/PD-1 blocking format. Furthermore, bispecific anti-PD-L1 sdAb-CD40L fusions that exhibited strong CD40 agonist activity was charactered within a PD-L1+ environment. Citation Format: Jean-Baptiste Marchand, Elodie Pintado, Marshall Dunlop, Christelle Remy, Patricia Kleinpeter, Shirley Shön, Fend Laetitia, Renée Brandely, Delphine Suhner, Eline Winter, Nathalie Silvestre, Claire Huguet, Peter Fitzgerald, Eric Quéméneur. Selection of an optimal anti-PD-L1 single domain antibody format for the vectorization into oncolytic vaccinia virus and the generation of bispecific immunomodulators [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1885.

A Deletion in Duchenne Muscular Dystrophy Gene Found Through Whole Exome Sequencing in Iran.

Duchenne muscular dystrophy (DMD) is a severe progressive X-linked neuromuscular illness that affects movement through mutations in dystrophin gene. The mutation leads to insufficient, lack of, or dysfunction of dystrophin. The cause of DMD was determined in an Iranian family. Exome sequencing was carried out along with a complete physical examination of the family. In silico methods were applied to find the alteration in the protein structure. The homozygous variant in DMD gene (NM-004006.2) was defined as c.2732-2733delTT (p.Phe911CysfsX8) in exon 21. In addition, phylogenetic conservation study of the human dystrophin protein sequence revealed that phenylalanine 911 is one of the evolutionarily conserved amino acids. In conclusion, our study indicated a new deletion in the DMD gene in the affected family. This deletion with an X-linked inheritance pattern is new in Iran. These findings could facilitate genetic counseling for this family and other patients in the future.

Comparative Analysis of Olfactory Receptor Repertoires Sheds Light on the Diet Adaptation of the Bamboo-Eating Giant Panda Based on the Chromosome-Level Genome

The giant panda (Ailuropoda melanoleuca) is the epitome of a flagship species for wildlife conservation and also an ideal model of adaptive evolution. As an obligate bamboo feeder, the giant panda relies on the olfaction for food recognition. The number of olfactory receptor (OR) genes and the rate of pseudogenes are the main factors affecting the olfactory ability of animals. In this study, we used the chromosome-level genome of the giant panda to identify OR genes and compared the genome sequences of OR genes with five other Ursidae species (spectacled bear (Tremarctos ornatus), American black bear (Ursus americanus), brown bear (Ursus arctos), polar bear (Ursus maritimus) and Asian black bear (Ursus thibetanus)). The giant panda had 639 OR genes, including 408 functional genes, 94 partial OR genes and 137 pseudogenes. Among them, 222 OR genes were detected and distributed on 18 chromosomes, and chromosome 8 had the most OR genes. A total of 448, 617, 582, 521 and 792 OR genes were identified in the spectacled bear, American black bear, brown bear, polar bear and Asian black bear, respectively. Clustering analysis based on the OR protein sequences of the six species showed that the OR genes distributed in 69 families and 438 subfamilies based on sequence similarity, and the six mammals shared 72 OR gene subfamilies, while the giant panda had 31 unique OR gene subfamilies (containing 35 genes). Among the 35 genes, there are 10 genes clustered into 8 clusters with 10 known human OR genes (OR8J3, OR51I1, OR10AC1, OR1S2, OR1S1, OR51S1, OR4M1, OR4M2, OR51T1 and OR5W2). However, the kind of odor molecules can be recognized by the 10 known human OR genes separately, which needs further research. The phylogenetic tree showed that 345 (about 84.56%) functional OR genes were clustered as Class-II, while only 63 (about 15.44%) functional OR genes were clustered as Class-I, which required further and more in-depth research. The potential odor specificity of some giant panda OR genes was identified through the similarity to human protein sequences. Sequences similar to OR2B1, OR10G3, OR11H6 and OR11H7P were giant panda-specific lacking, which may be related to the transformation and specialization from carnivore to herbivore of the giant panda. Since our reference to flavoring agents comes from human research, the possible flavoring agents from giant panda-specific OR genes need further investigation. Moreover, the conserved motifs of OR genes were highly conserved in Ursidae species. This systematic study of OR genes in the giant panda will provide a solid foundation for further research on the olfactory function and variation of the giant panda.

Improved global protein homolog detection with major gains in function identification.

There are several hundred million protein sequences, but the relationships among them are not fully available from existing homolog detection methods. There is an essential need for an improved method to push homolog detection to lower levels of sequence identity. The method used here relies on a language model to represent proteins numerically in a matrix (an embedding) and uses discrete cosine transforms to compress the data to extract the most essential part, significantly reducing the data size. This PRotein Ortholog Search Tool (PROST) is significantly faster with linear runtimes, and most importantly, computes the distances between pairs of protein sequences to yield homologs at significantly lower levels of sequence identity than previously. The extent of allosteric effects in proteins points out the importance of global aspects of structure and sequence. PROST excels at global homology detection but not at detecting local homologs. Results are validated by strong similarities between the corresponding pairs of structures. The number of remote homologs detected increased significantly and pushes the effective sequence matches more deeply into the twilight zone. Human protein sequences presently having no assigned function now find significant numbers of putative homologs for 93% of cases and structurally verified assigned functions for 76.4% of these cases. The data compression enables massive searches for homologs with short search times while yielding significant gains in the numbers of remote homologs detected. The method is sufficiently efficient to permit whole-genome/proteome comparisons. The PROST web server is accessible at https://mesihk.github.io/prost.

Protein Interaction Prediction Method Based on Feature Engineering and XGBoost

Human protein interaction prediction studies occupy an important place in systems biology. The understanding of human protein interaction networks and interactome will provide important insights into the regulation of developmental, physiological and pathological processes. In this study, we propose a method based on feature engineering and integrated learning algorithms to construct protein interaction prediction models. Principal Component Analysis (PCA) and Locally Linear Embedding (LLE) dimensionality reduction methods were used to extract sequence features from the 174-dimensional human protein sequence vector after Normalized Difference Sequence Feature (NDSF) encoding, respectively. The classification performance of three integrated learning methods (AdaBoost, Extratrees, XGBoost) applied to PCA and LLE features was compared, and the best combination of parameters was found using cross-validation and grid search methods. The results show that the classification accuracy is significantly higher when using the linear dimensionality reduction method PCA than the nonlinear dimensionality reduction method LLE. the classification with XGBoost achieves a model accuracy of 99.2%, which is the best performance among all models. This study suggests that NDSF combined with PCA and XGBoost may be an effective strategy for classifying different human protein interactions.

Improved Coverage of the N-Terminome by Combining ChaFRADIC with Alternative Proteases.

Many proteolytic cleavage events cannot be covered with conventional trypsin-based N-terminomics workflows. These typically involve the derivatization of protein N-termini and Lys residues as an initial step, such that trypsin will cleave C-terminal of arginine but not lysine residues (ArgC-like cleavage). From 20,422 reviewed human protein sequences in Uniprot, 3597 have known N-terminal signal peptides. An in silico ArgC-like digestion of the corresponding 3597 mature protein sequences reveals that-even for these well-known and well-studied proteolytic events-trypsin-based N-terminomics workflows may miss up to 50% of signaling cleavage events as the corresponding neo-N-terminal peptides will have an unfavorable length of <7 (875 peptides) or >30 (911 peptides) amino acids. In this chapter, we provide a protocol that can be applied to all kinds of samples to improve access to this "inaccessible" N-terminome, by making use of the alternative, broad-specificity protease subtilisin for fast and reproducible digestion of proteins.

Shared 6mer Peptides of Human and Omicron (21K and 21L) at SARS-CoV-2 Mutation Sites.

We investigated the short sequences involving Omicron 21K and Omicron 21L variants to reveal any possible molecular mimicry-associated autoimmunity risks and changes in those. We first identified common 6mers of the viral and human protein sequences present for both the mutant (Omicron) and nonmutant (SARS-CoV-2) versions of the same viral sequence and then predicted the binding affinities of those sequences to the HLA supertype representatives. We evaluated change in the potential autoimmunity risk, through comparative assessment of the nonmutant and mutant viral sequences and their similar human peptides with common 6mers and affinities to the same HLA allele. This change is the lost and the new, or de novo, autoimmunity risk, associated with the mutations in the Omicron 21K and Omicron 21L variants. Accordingly, e.g., the affinity of virus-similar sequences of the Ig heavy chain junction regions shifted from the HLA-B*15:01 to the HLA-A*01:01 allele at the mutant sequences. Additionally, peptides of different human proteins sharing 6mers with SARS-CoV-2 proteins at the mutation sites of interest and with affinities to the HLA-B*07:02 allele, such as the respective SARS-CoV-2 sequences, were lost. Among all, any possible molecular mimicry-associated novel risk appeared to be prominent in HLA-A*24:02 and HLA-B*27:05 serotypes upon infection with Omicron 21L. Associated disease, pathway, and tissue expression data supported possible new risks for the HLA-B*27:05 and HLA-A*01:01 serotypes, while the risks for the HLA-B*07:02 serotypes could have been lost or diminished, and those for the HLA-A*03:01 serotypes could have been retained, for the individuals infected with Omicron variants under study. These are likely to affect the complications related to cross-reactions influencing the relevant HLA serotypes upon infection with Omicron 21K and Omicron 21L.

Genomic Analysis of S protein in SARS-CoV-2 samples from Washington D.C

COVID-19, otherwise known as the coronavirus, ranks as one of the top ten world’s worst epidemics since recorded history. As severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) evolves, it accumulates changes in the genetic code, commonly known as mutations. The greater spreadability of the SARS-CoV-2 variants is hypothesized to be driven by key mutations that are distinguished by higher mutation frequency and mutation density. This study investigates variants of the SARS-CoV-2 virus and identifies mutations that are of particular importance for future antibody, vaccine, and drug development. The human spike protein sequences from Washington D.C were analyzed by being compared to the reference SARS-COV-2 protein sequence from Wuhan-Hu-1. Out of 7012 sequences analyzed, a total of 2463 distinct mutations, at 1251 sites were observed. The mutations observed in the spike proteins are discussed in the context of mutation frequency, characteristics of mutations at different regions, number of mutations per sample, and mutation density at different regions. Different variants observed in the sequences, and how they have adapted from the existing variants such as Alpha, Delta, and Omicron are also discussed in this research paper. This study summarizes the mutation characteristics that influenced the virus' advantage and identifies the mutations to be considered for future antibody, vaccine, and drug development against the virus.

MLysPRED: graph-based multi-view clustering and multi-dimensional normal distribution resampling techniques to predict multiple lysine sites.

Posttranslational modification of lysine residues, K-PTM, is one of the most popular PTMs. Some lysine residues in proteins can be continuously or cascaded covalently modified, such as acetylation, crotonylation, methylation and succinylation modification. The covalent modification of lysine residues may have some special functions in basic research and drug development. Although many computational methods have been developed to predict lysine PTMs, up to now, the K-PTM prediction methods have been modeled and learned a single class of K-PTM modification. In view of this, this study aims to fill this gap by building a multi-label computational model that can be directly used to predict multiple K-PTMs in proteins. In this study, a multi-label prediction model, MLysPRED, is proposed to identify multiple lysine sites using features generated from human protein sequences. In MLysPRED, three kinds of multi-label sequence encoding algorithms (MLDBPB, MLPSDAAP, MLPSTAAP) are proposed and combined with three encoding strategies (CHHAA, DR and Kmer) to convert preprocessed lysine sequences into effective numerical features. A multidimensional normal distribution oversampling technique and graph-based multi-view clustering under-sampling algorithm were first proposed and incorporated to reduce the proportion of the original training samples, and multi-label nearest neighbor algorithm is used for classification. It is observed that MLysPRED achieved an Aiming of 92.21%, Coverage of 94.98%, Accuracy of 89.63%, Absolute-True of 81.46% and Absolute-False of 0.0682 on the independent datasets. Additionally, comparison of results with five existing predictors also indicated that MLysPRED is very promising and encouraging to predict multiple K-PTMs in proteins. For the convenience of the experimental scientists, 'MLysPRED' has been deployed as a user-friendly web-server at http://47.100.136.41:8181.

Integration of Human Protein Sequence and Protein-Protein Interaction Data by Graph Autoencoder to Identify Novel Protein-Abnormal Phenotype Associations.

Understanding gene functions and their associated abnormal phenotypes is crucial in the prevention, diagnosis and treatment against diseases. The Human Phenotype Ontology (HPO) is a standardized vocabulary for describing the phenotype abnormalities associated with human diseases. However, the current HPO annotations are far from completion, and only a small fraction of human protein-coding genes has HPO annotations. Thus, it is necessary to predict protein-phenotype associations using computational methods. Protein sequences can indicate the structure and function of the proteins, and interacting proteins are more likely to have same function. It is promising to integrate these features for predicting HPO annotations of human protein. We developed GraphPheno, a semi-supervised method based on graph autoencoders, which does not require feature engineering to capture deep features from protein sequences, while also taking into account the topological properties in the protein–protein interaction network to predict the relationships between human genes/proteins and abnormal phenotypes. Cross validation and independent dataset tests show that GraphPheno has satisfactory prediction performance. The algorithm is further confirmed on automatic HPO annotation for no-knowledge proteins under the benchmark of the second Critical Assessment of Functional Annotation, 2013–2014 (CAFA2), where GraphPheno surpasses most existing methods. Further bioinformatics analysis shows that predicted certain phenotype-associated genes using GraphPheno share similar biological properties with known ones. In a case study on the phenotype of abnormality of mitochondrial respiratory chain, top prioritized genes are validated by recent papers. We believe that GraphPheno will help to reveal more associations between genes and phenotypes, and contribute to the discovery of drug targets.

Abstract 2854: Logic gates controlled by priming receptors increase specificity and potency of CAR T cells

Abstract CAR T cell therapies for solid tumors are limited by a paucity of tumor-specific single target antigens and insufficient potency due to limitations of CAR T biology. Logic gated (LG) CARs increase control of cell therapy activation and target cell killing by requiring recognition of two or more antigens through AND or AND-NOT Boolean logic. They represent an attractive strategy for cell therapy to improve the therapeutic index by increasing specificity for tumor antigens. We created an AND logic gate cassette deliverable by CRISPR-mediated, non-viral, site-specific integration into human T cells. This logic gate activates expression of a MSLN-targeting CAR upon ALPP/G binding of a priming receptor (PrimeR) within high-grade serous carcinoma ovarian tumors. The PrimeR triggers proteolytic release of a chimeric, fully human transcription factor, which then translocates to the nucleus to induce expression of a minigene encoding a CAR. The PrimeR is built only from human protein sequences, reducing theoretical risk of cell therapy immunogenicity and rejection within a patient. In order to test the specificity of our LG CARs, we integrated LG cassettes into a defined site in the T cell genome, and then co-cultured T cells with K562 cells engineered with only one target antigen (K562-ALPG or K562-MSLN), both target antigens (K562-ALPG/MSLN), or neither target antigen (K562). LG CAR T cells only killed K562-ALPG/MSLN cells. The specificity of this LG was also demonstrated in vivo in a dual-flank K562 model, in which only the ALPG/MSLN tumor was inhibited by LG CAR T cells. CAR expression was not observed on LG CAR T cells recovered outside the ALPG/MSLN tumor, and no growth inhibition was observed in K562-MSLN tumors, unlike a constitutive CAR T control. To model priming antigen heterogeneity expected in tumors, we mixed K562-MSLN cells with varying proportions of K562-ALPG/MSLN cells. We found that the presence of only 5-15% ALPG/MSLN cells was sufficient to effect killing of all target cells by the induced MSLN CAR. The ALPG/MSLN LG CAR will be tested as a component of AB-1015, a novel therapy for indications including ovarian, fallopian tube, or primary peritoneal cancer. Citation Format: Jasper Williams, Kevin Martinez, Angela Boroughs, Laura Lim, Marian Sandoval, Cate Sue, Matthew Drever, Anzhi Yao, Joseph Choe, Maxim Sidorov, Sophia Phillips, Dina Polyak, Suchismita Mohanty, Stephen Santoro, Aaron Cooper, W. Nicholas Haining. Logic gates controlled by priming receptors increase specificity and potency of CAR T cells [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2854.