Geoffrey Henry Siwo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Geoffrey Henry Siwo has a total of 91 co-authors. Here are the top co-authors by number of shared publications:
Nitesh Chawla
University of Notre Dame
3 shared publicationsAkbar K. Waljee
Department of Learning Health Sciences, University of Michigan Medical School
2 shared publications- Akbar K. Waljee
Center for Global Health Equity, Michigan Medicine, University of Michigan, Ann Arbor, Michigan, USA
2 shared publications - Akpéli V. Nordor
Epigene Labs
2 shared publications Amit G. Singal
Department of Internal Medicine, UT Southwestern Medical Center
2 shared publicationsAsako Tan
Eck Institute for Global Health, University of Notre Dame, Notre Dame, IN, USA; Department of Biological Sciences, University of Notre Dame, Notre Dame, IN, USA; Epicentre, Madison, WI, USA
2 shared publications
Latest Publications
- Open Access
Protein co-scientist aided simulations of the structural and functional impact of O-GlcNAc Transferase TPR domain deletions and amino acid substitutions associated with X-Linked Intellectual Disability
O-GlcNAc transferase (OGT) is an essential glycosyltransferase that regulates cellular processes through the addition of O-linked β-N-acetyglucosamine (O-GlcNAc) to target proteins. Substrate recognition and protein-protein interactions are mediated by the tetratricopeptide repeat (TPR) domain, yet the contributions of individual TPR repeats to structural stability and function remain poorly understood. Here, we employed an integrated computational framework combining an AI protein co-scientist, Speak to a Protein, which incorporates Boltz-2 for structure prediction and rDock for ligand binding evaluation, with external tools including AlphaFold3 for more advanced structure prediction and the HDOCK server for protein-protein docking. This framework was used to investigate how deletions of TPR domains and disease-associated amino acid substitutions within the TPR domain affect OGT structure and function. Five synthetic truncation OGT isoforms, thirteen single-TPR deletion variants, and five X-linked intellectual disability (XLID)-associated amino acid substitutions were modeled and aligned to wild-type OGT for comparison. Structural perturbations were quantified using backbone root mean square deviation (RMSD), while functional effects were assessed through docking analyses with UDP-GlcNAc, Host Cell Factor 1 (HCF1), and Tet methylcytosine dioxygenase 1 (TET1). Variation in TPR length altered global protein structure and induced long-range conformational changes within the catalytic domain despite no sequence modifications in that region. Single TPR deletions showed that individual repeats contribute unequally to structural stability, with deletions in mid-to-late TPR domains producing the largest structural disruptions. Disease-associated amino acid substitutions also generated conformational changes and reduced predicted ligand binding, consistent with known reductions in enzymatic activity in XLID. Protein-protein docking analyses demonstrated substrate-specific effects, as domain deletion and amino acid substitution variants of OGT had different interactions with TET1 and HCF1. Collectively, these findings identify the TPR domain as a regulatory domain that governs structure, ligand recognition, and protein interactions. More broadly, this work demonstrates how the integration of a protein co-scientist with protein structure modeling tools can be used to predict the structural and functional consequences of domain deletions or natural amino acid variants.
Open AccessProtein co-scientist aided simulations of the structural and functional impact of O-GlcNAc Transferase TPR domain deletions and amino acid substitutions associated with X-Linked Intellectual Disability
O-GlcNAc transferase (OGT) is an essential glycosyltransferase that regulates cellular processes through the addition of O-linked β-N-acetyglucosamine (O-GlcNAc) to target proteins. Substrate recognition and protein-protein interactions are mediated by the tetratricopeptide repeat (TPR) domain, yet the contributions of individual TPR repeats to structural stability and function remain poorly understood. Here, we employed an integrated computational framework combining an AI protein co-scientist, Speak to a Protein, which incorporates Boltz-2 for structure prediction and rDock for ligand binding evaluation, with external tools including AlphaFold3 for more advanced structure prediction and the HDOCK server for protein-protein docking. This framework was used to investigate how deletions of TPR domains and disease-associated amino acid substitutions within the TPR domain affect OGT structure and function. Five synthetic truncation OGT isoforms, thirteen single-TPR deletion variants, and five X-linked intellectual disability (XLID)-associated amino acid substitutions were modeled and aligned to wild-type OGT for comparison. Structural perturbations were quantified using backbone root mean square deviation (RMSD), while functional effects were assessed through docking analyses with UDP-GlcNAc, Host Cell Factor 1 (HCF1), and Tet methylcytosine dioxygenase 1 (TET1). Variation in TPR length altered global protein structure and induced long-range conformational changes within the catalytic domain despite no sequence modifications in that region. Single TPR deletions showed that individual repeats contribute unequally to structural stability, with deletions in mid-to-late TPR domains producing the largest structural disruptions. Disease-associated amino acid substitutions also generated conformational changes and reduced predicted ligand binding, consistent with known reductions in enzymatic activity in XLID. Protein-protein docking analyses demonstrated substrate-specific effects, as domain deletion and amino acid substitution variants of OGT had different interactions with TET1 and HCF1. Collectively, these findings identify the TPR domain as a regulatory domain that governs structure, ligand recognition, and protein interactions. More broadly, this work demonstrates how the integration of a protein co-scientist with protein structure modeling tools can be used to predict the structural and functional consequences of domain deletions or natural amino acid variants.
- Open Access
- 2 October 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (10)
Assessment of LLMs on basic bioinformatics tasks requiring multi-step reasoning and code writing
Motivation: Large Language Models (LLMs), AI agents and co-scientists promise to accelerate scientific discovery across fields ranging from chemistry to biology. Bioinformatics—the analysis of DNA, RNA and protein sequences plays a crucial role in biological research and is especially amenable to AI-driven automation given its computational nature. Here, we assess the bioinformatics capabilities of popular general-purpose LLMs on a set of tasks covering basic analytical questions that include code writing and multi-step reasoning in the domain. Results: Utilizing questions from Rosalind, a bioinformatics educational platform, we compare the performance of the LLMs on 104 questions relative to expert ground truth and 68 760 humans with varying levels of expertise. Among the frontier LLMs evaluated (GPT-5.2, GPT-4o, GPT-3.5, Llama-3-70B, Llama-4 and Claude Sonnet 4.6), GPT-5.2 exhibited top performance, providing correct answers for 80/104 (77%) questions followed by Claude Sonnet 4.6. LLM performance varied widely depending on the biological categories of questions. While the latest generation of frontier LLM models showed overall improvements in performance, their performance in specific categories of biological questions was lower than first generation models. Overall, LLMs performance mirrored that of humans with lower performance in tasks in which humans had low performance and vice versa. However, LLMs also failed in some instances where most humans were correct and, in a few cases, LLMs excelled where most humans failed. To the best of our knowledge, this presents the first assessment of general purpose LLMs on basic bioinformatics tasks in distinct areas relative to both expert and non-expert ground truths. Availability and Implementation: The complete analysis was implemented in Python; all code is available at https://github.com/SiwoResearch/lllm-bioinformatics.
- 2 October 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (10)
Open AccessAssessment of LLMs on basic bioinformatics tasks requiring multi-step reasoning and code writing
Motivation: Large Language Models (LLMs), AI agents and co-scientists promise to accelerate scientific discovery across fields ranging from chemistry to biology. Bioinformatics—the analysis of DNA, RNA and protein sequences plays a crucial role in biological research and is especially amenable to AI-driven automation given its computational nature. Here, we assess the bioinformatics capabilities of popular general-purpose LLMs on a set of tasks covering basic analytical questions that include code writing and multi-step reasoning in the domain. Results: Utilizing questions from Rosalind, a bioinformatics educational platform, we compare the performance of the LLMs on 104 questions relative to expert ground truth and 68 760 humans with varying levels of expertise. Among the frontier LLMs evaluated (GPT-5.2, GPT-4o, GPT-3.5, Llama-3-70B, Llama-4 and Claude Sonnet 4.6), GPT-5.2 exhibited top performance, providing correct answers for 80/104 (77%) questions followed by Claude Sonnet 4.6. LLM performance varied widely depending on the biological categories of questions. While the latest generation of frontier LLM models showed overall improvements in performance, their performance in specific categories of biological questions was lower than first generation models. Overall, LLMs performance mirrored that of humans with lower performance in tasks in which humans had low performance and vice versa. However, LLMs also failed in some instances where most humans were correct and, in a few cases, LLMs excelled where most humans failed. To the best of our knowledge, this presents the first assessment of general purpose LLMs on basic bioinformatics tasks in distinct areas relative to both expert and non-expert ground truths. Availability and Implementation: The complete analysis was implemented in Python; all code is available at https://github.com/SiwoResearch/lllm-bioinformatics.
- Open Access
Development and Evaluation of Artificial Intelligence–Assisted Decision Support System for Public Health Emergency Classification and Escalation in Kenya
Background: Timely assessment, classification, and escalation of public health events are essential for effective outbreak response, yet decision-making after event detection remains challenging because of fragmented guidance and variable interpretation of escalation criteria.To strengthen public health emergency management, Kenya developed the Decision-Making Tool for Public Health Emergencies (DMT-PHE), a framework for event assessment, classification, notification, and escalation. An artificial intelligence (AI)-enabled version, the DMT-PHE AI Agent, was subsequently developed to operationalize the framework through decision support. This study describes the development of the DMT-PHE AI Agent and evaluates its performance, usability, safety, and user acceptability. Methods: The DMT-PHE AI Agent was developed using a retrieval-augmented generation architecture supported by a curated knowledge base derived from the validated DMT-PHE framework and related public health guidance. A simulation-based pilot evaluation was conducted among 11 public health professionals who independently assessed three standardized outbreak scenarios. AI-generated recommendations were compared with expert-defined gold standards. Outcomes included concordance, response-action coverage, citation performance, safety, usability, and user acceptability. Results: Thirty-three scenario evaluations were completed. The AI Agent achieved an overall weighted concordance score of 0.924, with exact agreement of 90.9% for Public Health Events of Initially Unknown Etiology, 81.8% for Rift Valley fever, and 90.9% for Mpox. Citation support was provided in 78.8% of interactions, with no incorrect citations or major safety concerns identified. The mean System Usability Scale score was 85.2, while participants reported high trust (4.27/5), contextual relevance (4.55/5), and perceived time savings (4.82/5). Conclusions: The DMT-PHE AI Agent demonstrated that a nationally validated public health emergency decision framework can be successfully translated into an AI-enabled decision-support system. These findings provide early evidence that AI can augment public health emergency decision-making by delivering structured, transparent, and context-specific recommendations while maintaining human oversight, offering a practical model for operationalizing national public health guidance.
Open AccessDevelopment and Evaluation of Artificial Intelligence–Assisted Decision Support System for Public Health Emergency Classification and Escalation in Kenya
Background: Timely assessment, classification, and escalation of public health events are essential for effective outbreak response, yet decision-making after event detection remains challenging because of fragmented guidance and variable interpretation of escalation criteria.To strengthen public health emergency management, Kenya developed the Decision-Making Tool for Public Health Emergencies (DMT-PHE), a framework for event assessment, classification, notification, and escalation. An artificial intelligence (AI)-enabled version, the DMT-PHE AI Agent, was subsequently developed to operationalize the framework through decision support. This study describes the development of the DMT-PHE AI Agent and evaluates its performance, usability, safety, and user acceptability. Methods: The DMT-PHE AI Agent was developed using a retrieval-augmented generation architecture supported by a curated knowledge base derived from the validated DMT-PHE framework and related public health guidance. A simulation-based pilot evaluation was conducted among 11 public health professionals who independently assessed three standardized outbreak scenarios. AI-generated recommendations were compared with expert-defined gold standards. Outcomes included concordance, response-action coverage, citation performance, safety, usability, and user acceptability. Results: Thirty-three scenario evaluations were completed. The AI Agent achieved an overall weighted concordance score of 0.924, with exact agreement of 90.9% for Public Health Events of Initially Unknown Etiology, 81.8% for Rift Valley fever, and 90.9% for Mpox. Citation support was provided in 78.8% of interactions, with no incorrect citations or major safety concerns identified. The mean System Usability Scale score was 85.2, while participants reported high trust (4.27/5), contextual relevance (4.55/5), and perceived time savings (4.82/5). Conclusions: The DMT-PHE AI Agent demonstrated that a nationally validated public health emergency decision framework can be successfully translated into an AI-enabled decision-support system. These findings provide early evidence that AI can augment public health emergency decision-making by delivering structured, transparent, and context-specific recommendations while maintaining human oversight, offering a practical model for operationalizing national public health guidance.
- Open Access
- 20 May 2026
- Published by Frontiers Media SA in Frontiers in Oncology
- Vol. 16, 1727389
Pan-cancer molecular signatures connecting aspartate transaminase to cancer prognosis, metabolic and immune signatures
Serum aspartate transaminase (sAST) level is used routinely in conjunction with other clinical assays to assess liver health and disease. Increasing evidence suggests that sAST is associated with all-cause mortality and has prognostic value in several cancers, including gastrointestinal and urothelial cancers. Here, we undertake a systems approach to unravel molecular connections between AST and cancer prognosis, metabolism, and immune signatures at the transcriptomic and proteomic levels. We find that GOT1 and GOT2 expression is reversed in tumors relative to normal tissues whereby tumors residing in tissues that normally have low expression tend to have high GOT1/GOT2 expression and vice versa. Expression of GOT1 and GOT2 is associated with overall survival in several tumors across distinct tissues. At the proteomic level, expression of AST is associated with distinct pan-cancer molecular subtypes with an enrichment of specific metabolic and immune signatures. The GOT1 interactome is enriched with the targets of cancer-associated miRNAs, specifically mir34a – a promising cancer therapeutic, while the GOT2 interactome is enriched with proteins that interact with cancer-associated transcription factors. Our findings suggest that perturbations in the levels of GOT1/GOT2 within specific tissues reflect pathophysiological changes beyond tissue damage and have implications for cancer metabolism, immune infiltration, prognosis, and treatment personalization.
- 20 May 2026
- Published by Frontiers Media SA in Frontiers in Oncology
- Vol. 16, 1727389
Open AccessPan-cancer molecular signatures connecting aspartate transaminase to cancer prognosis, metabolic and immune signatures
Serum aspartate transaminase (sAST) level is used routinely in conjunction with other clinical assays to assess liver health and disease. Increasing evidence suggests that sAST is associated with all-cause mortality and has prognostic value in several cancers, including gastrointestinal and urothelial cancers. Here, we undertake a systems approach to unravel molecular connections between AST and cancer prognosis, metabolism, and immune signatures at the transcriptomic and proteomic levels. We find that GOT1 and GOT2 expression is reversed in tumors relative to normal tissues whereby tumors residing in tissues that normally have low expression tend to have high GOT1/GOT2 expression and vice versa. Expression of GOT1 and GOT2 is associated with overall survival in several tumors across distinct tissues. At the proteomic level, expression of AST is associated with distinct pan-cancer molecular subtypes with an enrichment of specific metabolic and immune signatures. The GOT1 interactome is enriched with the targets of cancer-associated miRNAs, specifically mir34a – a promising cancer therapeutic, while the GOT2 interactome is enriched with proteins that interact with cancer-associated transcription factors. Our findings suggest that perturbations in the levels of GOT1/GOT2 within specific tissues reflect pathophysiological changes beyond tissue damage and have implications for cancer metabolism, immune infiltration, prognosis, and treatment personalization.
- Open Access
- 15 November 2024
- Published by Wolters Kluwer Health in Hepatology Communications
- Vol. 8 (12)
Machine-learning methodologies to predict disease progression in chronic hepatitis B in Africa
Background: Little is known about the determinants of disease progression among African patients with chronic HBV infection. Methods: We used machine-learning models with longitudinal data to establish predictive algorithms in a well-characterized cohort of Ethiopian HBV-infected patients without baseline liver fibrosis. Disease progression was defined as an increase in liver stiffness to >7.9 kPa or initiation of treatment based on meeting the eligibility criteria. Results: Twenty-four of 551 patients (4.4%) experienced disease progression after a median follow-up time of 69 months. A random forest model based on a combination of available laboratory tests (standard hematology and biochemistry) demonstrated the best predictive properties with the AUROC ranging from 0.82 to 0.88. Conclusion: We conclude that combined metrics based on simple and available laboratory tests had good predictive properties and should be explored further in larger HBV cohorts.
- 15 November 2024
- Published by Wolters Kluwer Health in Hepatology Communications
- Vol. 8 (12)
Open AccessMachine-learning methodologies to predict disease progression in chronic hepatitis B in Africa
Background: Little is known about the determinants of disease progression among African patients with chronic HBV infection. Methods: We used machine-learning models with longitudinal data to establish predictive algorithms in a well-characterized cohort of Ethiopian HBV-infected patients without baseline liver fibrosis. Disease progression was defined as an increase in liver stiffness to >7.9 kPa or initiation of treatment based on meeting the eligibility criteria. Results: Twenty-four of 551 patients (4.4%) experienced disease progression after a median follow-up time of 69 months. A random forest model based on a combination of available laboratory tests (standard hematology and biochemistry) demonstrated the best predictive properties with the AUROC ranging from 0.82 to 0.88. Conclusion: We conclude that combined metrics based on simple and available laboratory tests had good predictive properties and should be explored further in larger HBV cohorts.
- Open Access
Pan-cancer molecular signatures connecting aspartate transaminase (AST) to cancer prognosis, metabolic and immune signatures
Background Serum aspartate transaminase (sAST) level is used routinely in conjunction with other clinical assays to assess liver health and disease. Increasing evidence suggests that sAST is associated with all-cause mortality and has prognostic value in several cancers, including gastrointestinal and urothelial cancers. Here, we undertake a systems approach to unravel molecular connections between AST and cancer prognosis, metabolism, and immune signatures at the transcriptomic and proteomic levels. Methods We mined public gene expression data across multiple normal and cancerous tissues using the Genotype Tissue Expression (GTEX) resource and The Cancer Genome Atlas (TCGA) to assess the expression of genes encoding AST isoenzymes (GOT1 and GOT2) and their association with disease prognosis and immune infiltration signatures across multiple tumors. We examined the associations between AST and previously reported pan-cancer molecular subtypes characterized by distinct metabolic and immune signatures. We analyzed human protein-protein interaction networks for interactions between GOT1 and GOT2 with cancer-associated proteins. Using public databases and protein-protein interaction networks, we determined whether the subset of proteins that interact with AST (GOT1 and GOT2 interactomes) are enriched with proteins associated with specific diseases, miRNAs and transcription factors. Results We show that AST transcript isoforms (GOT1 and GOT2) are expressed across a wide range of normal tissues. AST isoforms are upregulated in tumors of the breast, lung, uterus, and thymus relative to normal tissues but downregulated in tumors of the liver, colon, brain, kidney and skeletal sarcomas. At the proteomic level, we find that the expression of AST is associated with distinct pan-cancer molecular subtypes with an enrichment of specific metabolic and immune signatures. Based on human protein-protein interaction data, AST physically interacts with multiple proteins involved in tumor initiation, suppression, progression, and treatment. We find enrichments in the AST interactomes for proteins associated with liver and lung cancer and dermatologic diseases. At the regulatory level, the GOT1 interactome is enriched with the targets of cancer-associated miRNAs, specifically mir34a – a promising cancer therapeutic, while the GOT2 interactome is enriched with proteins that interact with cancer-associated transcription factors. Conclusions Our findings suggest that perturbations in the levels of AST within specific tissues reflect pathophysiological changes beyond tissue damage and have implications for cancer metabolism, immune infiltration, prognosis, and treatment personalization.
Open AccessPan-cancer molecular signatures connecting aspartate transaminase (AST) to cancer prognosis, metabolic and immune signatures
Background Serum aspartate transaminase (sAST) level is used routinely in conjunction with other clinical assays to assess liver health and disease. Increasing evidence suggests that sAST is associated with all-cause mortality and has prognostic value in several cancers, including gastrointestinal and urothelial cancers. Here, we undertake a systems approach to unravel molecular connections between AST and cancer prognosis, metabolism, and immune signatures at the transcriptomic and proteomic levels. Methods We mined public gene expression data across multiple normal and cancerous tissues using the Genotype Tissue Expression (GTEX) resource and The Cancer Genome Atlas (TCGA) to assess the expression of genes encoding AST isoenzymes (GOT1 and GOT2) and their association with disease prognosis and immune infiltration signatures across multiple tumors. We examined the associations between AST and previously reported pan-cancer molecular subtypes characterized by distinct metabolic and immune signatures. We analyzed human protein-protein interaction networks for interactions between GOT1 and GOT2 with cancer-associated proteins. Using public databases and protein-protein interaction networks, we determined whether the subset of proteins that interact with AST (GOT1 and GOT2 interactomes) are enriched with proteins associated with specific diseases, miRNAs and transcription factors. Results We show that AST transcript isoforms (GOT1 and GOT2) are expressed across a wide range of normal tissues. AST isoforms are upregulated in tumors of the breast, lung, uterus, and thymus relative to normal tissues but downregulated in tumors of the liver, colon, brain, kidney and skeletal sarcomas. At the proteomic level, we find that the expression of AST is associated with distinct pan-cancer molecular subtypes with an enrichment of specific metabolic and immune signatures. Based on human protein-protein interaction data, AST physically interacts with multiple proteins involved in tumor initiation, suppression, progression, and treatment. We find enrichments in the AST interactomes for proteins associated with liver and lung cancer and dermatologic diseases. At the regulatory level, the GOT1 interactome is enriched with the targets of cancer-associated miRNAs, specifically mir34a – a promising cancer therapeutic, while the GOT2 interactome is enriched with proteins that interact with cancer-associated transcription factors. Conclusions Our findings suggest that perturbations in the levels of AST within specific tissues reflect pathophysiological changes beyond tissue damage and have implications for cancer metabolism, immune infiltration, prognosis, and treatment personalization.
- Open Access
Genomes contain relics of a triplet code connecting the origins of primordial RNA synthesis to the origins of genetically coded protein synthesis
Life on earth relies on three types of information polymers-DNA, RNA and proteins. In all organisms and viruses, these molecules are synthesized by the copying of pre-existing templates. A triplet-based code known as the genetic code guides the synthesis of proteins by complex enzymatic machines that decode genetic information in RNA sequences. The origin of the genetic code is one of the most fundamental questions in biology. In this study, computational analysis of about 5,000 species level metagenomes using techniques for the analysis of human language suggests that the genomes of extant organisms contain relics of a distinct triplet code that potentially predates the genetic code. This code defines the relationship between adjacent triplets in DNA/RNA sequences, whereby these triplets predominantly differ by a single base. Furthermore, adjacent triplets encode amino acids that are thought to have emerged around the same period in the earth’s early history. The results suggest that the order of triplets in primordial RNA sequences was associated with the availability of specific amino acids, perhaps due to a coupling of a triplet-based primordial RNA synthesis mechanism to a primitive mechanism of peptide bond formation. Together, this coupling could have given rise to early nucleic acid sequences and a system for encoding amino acid sequences in RNA, i.e. the genetic code. Thus, the central role of triplets in biology potentially extends to the primordial world, contributing to both the origins of genomes and the origins of genetically coded protein synthesis. Significance One of the most intriguing discoveries in biology is that the order of amino acids in each protein is determined by the order of nucleotides (commonly represented by the letters A, U, G, C) in a biological molecule known as RNA. The genetic code serves as a dictionary that maps each of the 64 triplets ‘words’ in RNA to the 20 amino acids, thereby specifying how information encoded in RNA is decoded into sequences of amino acids (i.e., proteins). The deciphering of the genetic code was one of the greatest discoveries of the 20th century (1968 Nobel Prize in Medicine and Physiology) and is central to modern molecular biology. Yet, how it came to be that the order of triplets in RNA encodes the sequence of the protein synthesized remains one of the most important enigmas of biology. Paradoxically, in all life forms proteins cannot be synthesized without RNA and RNA itself cannot also be synthesized without proteins, presenting a chicken and egg dilemma. By analyzing thousands of microbial genomes using approaches drawn from the field of natural language processing, this study finds that the order of triplets across genomes contains relics of an ancient triplet code, distinct from but closely connected to the genetic code. Unlike the genetic code which specifies the relationship between information in RNA and the sequence of proteins, this ancient code describes the relationship between adjacent triplets in extant genome sequences, whereby such triplets are often different from each other by a single letter. Triplets that are closely related by this ancient code encode amino acids that are thought to have emerged around the same period in the earth’s early history. In other words, a fossil record of the chronological order of appearance of amino acids on early earth appears written in genome sequences. This potentially demonstrates that the process by which RNA sequences were synthesized in the primordial world relied on triplets and was coupled to amino acids available at the time. Hence, the connections between primordial RNA synthesis and a primitive mechanism for linking amino acids to form peptides could have enabled one type of molecule (RNA) to code for the other (protein), facilitating the emergence of the genetic code.
Open AccessGenomes contain relics of a triplet code connecting the origins of primordial RNA synthesis to the origins of genetically coded protein synthesis
Life on earth relies on three types of information polymers-DNA, RNA and proteins. In all organisms and viruses, these molecules are synthesized by the copying of pre-existing templates. A triplet-based code known as the genetic code guides the synthesis of proteins by complex enzymatic machines that decode genetic information in RNA sequences. The origin of the genetic code is one of the most fundamental questions in biology. In this study, computational analysis of about 5,000 species level metagenomes using techniques for the analysis of human language suggests that the genomes of extant organisms contain relics of a distinct triplet code that potentially predates the genetic code. This code defines the relationship between adjacent triplets in DNA/RNA sequences, whereby these triplets predominantly differ by a single base. Furthermore, adjacent triplets encode amino acids that are thought to have emerged around the same period in the earth’s early history. The results suggest that the order of triplets in primordial RNA sequences was associated with the availability of specific amino acids, perhaps due to a coupling of a triplet-based primordial RNA synthesis mechanism to a primitive mechanism of peptide bond formation. Together, this coupling could have given rise to early nucleic acid sequences and a system for encoding amino acid sequences in RNA, i.e. the genetic code. Thus, the central role of triplets in biology potentially extends to the primordial world, contributing to both the origins of genomes and the origins of genetically coded protein synthesis. Significance One of the most intriguing discoveries in biology is that the order of amino acids in each protein is determined by the order of nucleotides (commonly represented by the letters A, U, G, C) in a biological molecule known as RNA. The genetic code serves as a dictionary that maps each of the 64 triplets ‘words’ in RNA to the 20 amino acids, thereby specifying how information encoded in RNA is decoded into sequences of amino acids (i.e., proteins). The deciphering of the genetic code was one of the greatest discoveries of the 20th century (1968 Nobel Prize in Medicine and Physiology) and is central to modern molecular biology. Yet, how it came to be that the order of triplets in RNA encodes the sequence of the protein synthesized remains one of the most important enigmas of biology. Paradoxically, in all life forms proteins cannot be synthesized without RNA and RNA itself cannot also be synthesized without proteins, presenting a chicken and egg dilemma. By analyzing thousands of microbial genomes using approaches drawn from the field of natural language processing, this study finds that the order of triplets across genomes contains relics of an ancient triplet code, distinct from but closely connected to the genetic code. Unlike the genetic code which specifies the relationship between information in RNA and the sequence of proteins, this ancient code describes the relationship between adjacent triplets in extant genome sequences, whereby such triplets are often different from each other by a single letter. Triplets that are closely related by this ancient code encode amino acids that are thought to have emerged around the same period in the earth’s early history. In other words, a fossil record of the chronological order of appearance of amino acids on early earth appears written in genome sequences. This potentially demonstrates that the process by which RNA sequences were synthesized in the primordial world relied on triplets and was coupled to amino acids available at the time. Hence, the connections between primordial RNA synthesis and a primitive mechanism for linking amino acids to form peptides could have enabled one type of molecule (RNA) to code for the other (protein), facilitating the emergence of the genetic code.
- Open Access
- 2 November 2021
- Published by Cambridge University Press (CUP) in ChemRxiv
Generative AI Design and Exploration of Nucleoside Analogs
Nucleosides are fundamental building blocks of DNA and RNA in all life forms and viruses. In addition, natural nucleosides and their analogs are critical in prebiotic chemistry, innate immunity, signaling, antiviral drug discovery and artificial synthesis of DNA / RNA sequences. Combined with the fact that quantitative structure activity relationships (QSAR) have been widely performed to understand their antiviral activity, nucleoside analogs could be used to benchmark generative molecular design. Here, we undertake the first generative design of nucleoside analogs using an approach that we refer to as the Conditional Randomized Transformer (CRT). We also benchmark our model against five previously published molecular generative models. We demonstrate that AI-generated molecules include nucleoside analogs that are of significance in a wide range of areas including prebiotic chemistry, antiviral drug discovery and synthesis of oligonucleotides. Our results show that CRT explores distinct molecular spaces and chemical transformations, some of which are similar to those undertaken by nature and medicinal chemists. Finally, we demonstrate the potential application of the CRT model in the generative design of molecules conditioned on Remdesivir and Molnupiravir as well as other nucleoside analogs with in vitro activity against SARS-CoV-2. One-Sentence Summary: Generative design of nucleoside analogs relevant to antiviral drug discovery, prebiotic chemistry and synthetic biology.
- 2 November 2021
- Published by Cambridge University Press (CUP) in ChemRxiv
Open AccessGenerative AI Design and Exploration of Nucleoside Analogs
Nucleosides are fundamental building blocks of DNA and RNA in all life forms and viruses. In addition, natural nucleosides and their analogs are critical in prebiotic chemistry, innate immunity, signaling, antiviral drug discovery and artificial synthesis of DNA / RNA sequences. Combined with the fact that quantitative structure activity relationships (QSAR) have been widely performed to understand their antiviral activity, nucleoside analogs could be used to benchmark generative molecular design. Here, we undertake the first generative design of nucleoside analogs using an approach that we refer to as the Conditional Randomized Transformer (CRT). We also benchmark our model against five previously published molecular generative models. We demonstrate that AI-generated molecules include nucleoside analogs that are of significance in a wide range of areas including prebiotic chemistry, antiviral drug discovery and synthesis of oligonucleotides. Our results show that CRT explores distinct molecular spaces and chemical transformations, some of which are similar to those undertaken by nature and medicinal chemists. Finally, we demonstrate the potential application of the CRT model in the generative design of molecules conditioned on Remdesivir and Molnupiravir as well as other nucleoside analogs with in vitro activity against SARS-CoV-2. One-Sentence Summary: Generative design of nucleoside analogs relevant to antiviral drug discovery, prebiotic chemistry and synthetic biology.
- Open Access
Molecular relationships between SARS-CoV-2 Spike protein and LIFR, a pneumonia protective IL-6 family cytokine receptor
The fine-tuned control of immune responses is attained by pairs of activating and inhibitory signaling receptors which modulate the quality and magnitude of immune responses. Some viruses exploit these pathways to enter host cells as well as interfere with immune responses. Here, we report that the SARS-CoV-1/2 Spike proteins (S) contain a potential inhibitory tyrosine- based immunoreceptor motif (ITIM) with the D614G variant occurring within this motif. Through an in silico screen of ITIM-containing human proteins, we find that the S-located ITIM is closely related to a previously reported ITIM in the cytoplasmic tail of the human Leukemia Inhibitory Factor Receptor (LIFR), a pneumonia protective IL-6 family cytokine receptor. To infer potential functional interactions between SARS-CoV-2 infection and LIFR expression, we performed single-cell transcriptome analysis of public datasets of lung tissues from healthy individuals and COVID-19 patients. We show that transcripts of LIFR and its ligand LIF are highly expressed in SARS-CoV-2 susceptible lung cells from mild and severe COVID-19 patients but not in healthy individuals. In addition, the human endogenous retroviral envelope gene ( ERVW-1 ) encoding a fusogenic protein of the same functional class as the S protein, is induced in SARS-CoV-2 susceptible cell subpopulations in COVID-19 patients with no detectable expression in healthy individuals. We also report that pulmonary epithelial cells express transcripts of several immunoreceptors including the ITIM-containing antibody receptor FCGR2B which is detectable in healthy and severe COVID-19 cases but not in mild cases. These results suggest that molecular dysregulation of ITIM-mediated inhibitory signaling by the SARS-CoV-2 S protein may play a role in COVID-19 immunopathology.
Open AccessMolecular relationships between SARS-CoV-2 Spike protein and LIFR, a pneumonia protective IL-6 family cytokine receptor
The fine-tuned control of immune responses is attained by pairs of activating and inhibitory signaling receptors which modulate the quality and magnitude of immune responses. Some viruses exploit these pathways to enter host cells as well as interfere with immune responses. Here, we report that the SARS-CoV-1/2 Spike proteins (S) contain a potential inhibitory tyrosine- based immunoreceptor motif (ITIM) with the D614G variant occurring within this motif. Through an in silico screen of ITIM-containing human proteins, we find that the S-located ITIM is closely related to a previously reported ITIM in the cytoplasmic tail of the human Leukemia Inhibitory Factor Receptor (LIFR), a pneumonia protective IL-6 family cytokine receptor. To infer potential functional interactions between SARS-CoV-2 infection and LIFR expression, we performed single-cell transcriptome analysis of public datasets of lung tissues from healthy individuals and COVID-19 patients. We show that transcripts of LIFR and its ligand LIF are highly expressed in SARS-CoV-2 susceptible lung cells from mild and severe COVID-19 patients but not in healthy individuals. In addition, the human endogenous retroviral envelope gene ( ERVW-1 ) encoding a fusogenic protein of the same functional class as the S protein, is induced in SARS-CoV-2 susceptible cell subpopulations in COVID-19 patients with no detectable expression in healthy individuals. We also report that pulmonary epithelial cells express transcripts of several immunoreceptors including the ITIM-containing antibody receptor FCGR2B which is detectable in healthy and severe COVID-19 cases but not in mild cases. These results suggest that molecular dysregulation of ITIM-mediated inhibitory signaling by the SARS-CoV-2 S protein may play a role in COVID-19 immunopathology.