Genotypic-phenotypic landscape computation based on first principle and deep learning.

The relationship between genotype and fitness is fundamental to evolution, but quantitatively mapping genotypes to fitness has remained challenging. We propose the Phenotypic-Embedding theorem (P-E theorem) that bridges genotype-phenotype through an encoder-decoder deep learning framework. Inspired by this, we proposed a more general first principle for correlating genotype-phenotype, and the P-E theorem provides a computable basis for the application of first principle. As an application example of the P-E theorem, we developed the Co-attention based Transformer model to bridge Genotype and Fitness model, a Transformer-based pre-train foundation model with downstream supervised fine-tuning that can accurately simulate the neutral evolution of viruses and predict immune escape mutations. Accordingly, following the calculation path of the P-E theorem, we accurately obtained the basic reproduction number (${R}_0$) of SARS-CoV-2 from first principles, quantitatively linked immune escape to viral fitness and plotted the genotype-fitness landscape. The theoretical system we established provides a general and interpretable method to construct genotype-phenotype landscapes, providing a new paradigm for studying theoretical and computational biology.

Off-target RNA mutation induced by DNA base editing and its elimination by mutagenesis.

Recently developed DNA base editing methods enable the direct generation of desired point mutations in genomic DNA without generating any double-strand breaks1-3, but the issue of off-target edits has limited the application of these methods. Although several previous studies have evaluated off-target mutations in genomic DNA4-8, it is now clear that the deaminases that are integral to commonly used DNA base editors often bind to RNA9-13. For example, the cytosine deaminase APOBEC1-which is used in cytosine base editors (CBEs)-targets both DNA and RNA12, and the adenine deaminase TadA-which is used in adenine base editors (ABEs)-induces site-specific inosine formation on RNA9,11. However, any potential RNA mutations caused by DNA base editors have not been evaluated. Adeno-associated viruses are the most common delivery system for gene therapies that involve DNA editing; these viruses can sustain long-term gene expression in vivo, so the extent of potential RNA mutations induced by DNA base editors is of great concern14-16. Here we quantitatively evaluated RNA single nucleotide variations (SNVs) that were induced by CBEs or ABEs. Both the cytosine base editor BE3 and the adenine base editor ABE7.10 generated tens of thousands of off-target RNA SNVs. Subsequently, by engineering deaminases, we found that three CBE variants and one ABE variant showed a reduction in off-target RNA SNVs to the baseline while maintaining efficient DNA on-target activity. This study reveals a previously overlooked aspect of off-target effects in DNA editing and also demonstrates that such effects can be eliminated by engineering deaminases.

SMDB: a Spatial Multimodal Data Browser.

Understanding the relationship between fine-scale spatial organization and biological function necessitates a tool that effectively combines spatial positions, morphological information, and spatial transcriptomics (ST) data. We introduce the Spatial Multimodal Data Browser (SMDB, https://www.biosino.org/smdb), a robust visualization web service for interactively exploring ST data. By integrating multimodal data, such as hematoxylin and eosin (H&E) images, gene expression-based molecular clusters, and more, SMDB facilitates the analysis of tissue composition through the dissociation of two-dimensional (2D) sections and the identification of gene expression-profiled boundaries. In a digital three-dimensional (3D) space, SMDB allows researchers to reconstruct morphology visualizations based on manually filtered spots or expand anatomical structures using high-resolution molecular subtypes. To enhance user experience, it offers customizable workspaces for interactive exploration of ST spots in tissues, providing features like smooth zooming, panning, 360-degree rotation in 3D and adjustable spot scaling. SMDB is particularly valuable in neuroscience and spatial histology studies, as it incorporates Allen's mouse brain anatomy atlas for reference in morphological research. This powerful tool provides a comprehensive and efficient solution for examining the intricate relationships between spatial morphology, and biological function in various tissues.

A Noncoding A-to-U Kozak Site Change Related to the High Transmissibility of Alpha, Delta, and Omicron VOCs.

Three prevalent SARS-CoV-2 variants of concern (VOCs) emerged and caused epidemic waves. It is essential to uncover advantageous mutations that cause the high transmissibility of VOCs. However, viral mutations are tightly linked, so traditional population genetic methods, including machine learning-based methods, cannot reliably detect mutations conferring a fitness advantage. In this study, we developed an approach based on the sequential occurrence order of mutations and the accelerated furcation rate in the pandemic-scale phylogenomic tree. We analyzed 3,777,753 high-quality SARS-CoV-2 genomic sequences and the epidemiology metadata using the Coronavirus GenBrowser. We found that two noncoding mutations at the same position (g.a28271-/u) may be crucial to the high transmissibility of Alpha, Delta, and Omicron VOCs although the noncoding mutations alone cannot increase viral transmissibility. Both mutations cause an A-to-U change at the core position -3 of the Kozak sequence of the N gene and significantly reduce the protein expression ratio of ORF9b to N. Using a convergent evolutionary analysis, we found that g.a28271-/u, S:p.P681H/R, and N:p.R203K/M occur independently on three VOC lineages, suggesting that coordinated changes of S, N, and ORF9b proteins are crucial to high viral transmissibility. Our results provide new insights into high viral transmissibility co-modulated by advantageous noncoding and nonsynonymous changes.

Donor-derived stem cell infusion for sustained pancytopenia after CD19 CAR-T therapy for relapsed patients post allogeneic stem cell transplantation.

OBJECTIVES: To investigate the effectiveness of donor-derived chimeric antigen receptor T (CAR-T) cells in the treatment of relapsed cases after allogeneic hematopoietic stem cell transplantation (allo-HSCT), and whether donor-derived peripheral blood stem cells (PBSCs) have a therapeutic effect on pancytopenia after CAR-T cell therapy. METHODS: We analyzed data from five adults with B-cell acute lymphoblastic leukemia (ALL) who had relapse after allo-HSCT and received donor-derived CAR-T cell therapy and donor-derived PBSCs to promote hematopoietic recovery. RESULTS: All patients had negative minimal residual disease after CAR-T therapy, grade 1-2 cytokine release syndrome, and developed grade 4 hematologic toxicity. During the pancytopenia stage after CAR-T cell therapy, donor-derived PBSCs were transfused without graft-versus-host disease (GVHD) prophylaxis. Four patients had grade I-II acute GVHD (aGVHD). After corticosteroid treatment, aGVHD resolved and hematopoiesis was restored. Although steroids in combination with etanercept and ruxolitinib relieved symptoms in one patient with grade IV aGVHD, complete hematopoietic recovery was not achieved, and the patient died due to severe infection. CONCLUSIONS: Donor-derived CAR-T cell therapy is safe and effective in patients with relapsed/refractory ALL after allo-HSCT. Donor-derived PBSCs infusion could achieve hematopoietic recovery with controllable aGVHD in patients with persistent pancytopenia.

PLRTE: Progressive learning for biomedical relation triplet extraction using large language models.

Document-level relation triplet extraction is crucial in biomedical text mining, aiding in drug discovery and the construction of biomedical knowledge graphs. Current language models face challenges in generalizing to unseen datasets and relation types in biomedical relation triplet extraction, which limits their effectiveness in these crucial tasks. To address this challenge, our study optimizes models from two critical dimensions: data-task relevance and granularity of relations, aiming to enhance their generalization capabilities significantly. We introduce a novel progressive learning strategy to obtain the PLRTE model. This strategy not only enhances the model's capability to comprehend diverse relation types in the biomedical domain but also implements a structured four-level progressive learning process through semantic relation augmentation, compositional instruction, and dual-axis level learning. Our experiments on the DDI and BC5CDR document-level biomedical relation triplet datasets demonstrate a significant performance improvement of 5% to 20% over the current state-of-the-art baselines. Furthermore, our model exhibits exceptional generalization capabilities on the unseen Chemprot and GDA datasets, further validating the effectiveness of optimizing data-task association and relation granularity for enhancing model generalizability.

A rationally engineered cytosine base editor retains high on-target activity while reducing both DNA and RNA off-target effects.

Cytosine base editors (CBEs) offer a powerful tool for correcting point mutations, yet their DNA and RNA off-target activities have caused concerns in biomedical applications. We describe screens of 23 rationally engineered CBE variants, which reveal mutation residues in the predicted DNA-binding site can dramatically decrease the Cas9-independent off-target effects. Furthermore, we obtained a CBE variant-YE1-BE3-FNLS-that retains high on-target editing efficiency while causing extremely low off-target edits and bystander edits.

Using Machine Learning Methods to Study Colorectal Cancer Tumor Micro-Environment and Its Biomarkers.

Colorectal cancer (CRC) is a leading cause of cancer deaths worldwide, and the identification of biomarkers can improve early detection and personalized treatment. In this study, RNA-seq data and gene chip data from TCGA and GEO were used to explore potential biomarkers for CRC. The SMOTE method was used to address class imbalance, and four feature selection algorithms (MCFS, Borota, mRMR, and LightGBM) were used to select genes from the gene expression matrix. Four machine learning algorithms (SVM, XGBoost, RF, and kNN) were then employed to obtain the optimal number of genes for model construction. Through interpretable machine learning (IML), co-predictive networks were generated to identify rules and uncover underlying relationships among the selected genes. Survival analysis revealed that INHBA, FNBP1, PDE9A, HIST1H2BG, and CADM3 were significantly correlated with prognosis in CRC patients. In addition, the CIBERSORT algorithm was used to investigate the proportion of immune cells in CRC tissues, and gene mutation rates for the five selected biomarkers were explored. The biomarkers identified in this study have significant implications for the development of personalized therapies and could ultimately lead to improved clinical outcomes for CRC patients.

DysRegSig: an R package for identifying gene dysregulations and building mechanistic signatures in cancer.

SUMMARY: Dysfunctional regulations of gene expression programs relevant to fundamental cell processes can drive carcinogenesis. Therefore, systematically identifying dysregulation events is an effective path for understanding carcinogenesis and provides insightful clues to build predictive signatures with mechanistic interpretability for cancer precision medicine. Here, we implemented a machine learning-based gene dysregulation analysis framework in an R package, DysRegSig, which is capable of exploring gene dysregulations from high-dimensional data and building mechanistic signature based on gene dysregulations. DysRegSig can serve as an easy-to-use tool to facilitate gene dysregulation analysis and follow-up analysis. AVAILABILITY AND IMPLEMENTATION: The source code and user's guide of DysRegSig are freely available at Github: https://github.com/SCBIT-YYLab/DysRegSig. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Associations of erythrocyte polyunsaturated fatty acids with incidence of stroke and stroke types in adult Chinese: a prospective study of over 8000 individuals.

PURPOSE: There is limited and inconsistent evidence about the relationships of erythrocyte polyunsaturated fatty acids (PUFAs) with stroke and stroke types, particularly in China where the stroke rates are high. We aimed to investigate the associations of different erythrocyte PUFAs with incidence of total stroke, ischemic stroke (IS), and intracerebral hemorrhage (ICH) in Chinese adults. METHODS: In the prospective China Kadoorie Biobank, erythrocyte PUFAs were measured using gas chromatography in 10,563 participants who attended 2013-14 resurvey. After a mean follow-up of 3.8 years, 412 incident stroke cases (342 IS, 53 ICH) were recorded among 8,159 participants without prior vascular diseases or diabetes. Cox regression yielded adjusted hazard ratios (HRs) for stroke associated with 13 PUFAs. RESULTS: Overall, the mean body mass index was 24.0 (3.4) kg/m2 and the mean age was 58.1 (9.9) years. In multivariable analyses, 18:2n-6 was positively associated with ICH (HR = 2.33 [95% CIs 1.41, 3.82] for top versus bottom quintile, Ptrend = 0.007), but inversely associated with IS (0.69 [0.53,0.90], Ptrend = 0.027), while 20:3n-6 was positively associated with risk of IS (1.64 [1.32,2.04], Ptrend < 0.001), but not with ICH. Inverted-U shape curve associations were observed of 20:5n-3 with IS (Pnonlinear = 0.002) and total stroke (Pnonlinear = 0.008), with a threshold at 0.70%. After further adjustment for conventional CVD risk factors and dietary factors, these associations remained similar. CONCLUSION: Among relatively lean Chinese adults, erythrocyte PUFAs 18:2n-6, 20:3n-6 and 20:5n-3 showed different associations with risks of IS and ICH. These results would improve the understanding of stroke etiology.

PhenoModifier: a genetic modifier database for elucidating the genetic basis of human phenotypic variation.

From clinical observations to large-scale sequencing studies, the phenotypic impact of genetic modifiers is evident. To better understand the full spectrum of the genetic contribution to human disease, concerted efforts are needed to construct a useful modifier resource for interpreting the information from sequencing data. Here, we present the PhenoModifier (https://www.biosino.org/PhenoModifier), a manually curated database that provides a comprehensive overview of human genetic modifiers. By manually curating over ten thousand published articles, 3078 records of modifier information were entered into the current version of PhenoModifier, related to 288 different disorders, 2126 genetic modifier variants and 843 distinct modifier genes. To help users probe further into the mechanism of their interested modifier genes, we extended the yeast genetic interaction data and yeast quantitative trait loci to the human and we also integrated GWAS data into the PhenoModifier to assist users in evaluating all possible phenotypes associated with a modifier allele. As the first comprehensive resource of human genetic modifiers, PhenoModifier provides a more complete spectrum of genetic factors contributing to human phenotypic variation. The portal has a broad scientific and clinical scope, spanning activities relevant to variant interpretation for research purposes as well as clinical decision making.

Improving the genome assembly of rabbits with long-read sequencing.

The European rabbit (Oryctolagus cuniculus) is important as a biomedical model given its unique features in immunity and metabolism. The current reference genome OryCun2.0 established with whole-genome shotgun sequencing was quite fragmented and had not been updated for ten years. In this work, we provided a new rabbit genome assembly UM_NZW_1.0 to improve OryCun2.0 by leveraging the contig lengths based on long-read sequencing and a wealth of available Illumina paired-end sequence data. UM_NZW_1.0 showed a remarkable increase of continuity compared with OryCun2.0, with 5 times longer contig N50 and approximately 75% gaps closed. Many of the closed gaps were overlapped with protein-coding genes or transcriptional features, resulting in an enhancement of gene annotations. In particular, UM_NZW_1.0 presented a more complete landscape of the MHC region and the IGH locus, therefore provided a valuable resource for future researches on rabbits.

Simultaneous zygotic inactivation of multiple genes in mouse through CRISPR/Cas9-mediated base editing.

In vivo genetic mutation has become a powerful tool for dissecting gene function; however, multi-gene interaction and the compensatory mechanisms involved can make findings from single mutations, at best difficult to interpret, and, at worst, misleading. Hence, it is necessary to establish an efficient way to disrupt multiple genes simultaneously. CRISPR/Cas9-mediated base editing disrupts gene function by converting a protein-coding sequence into a stop codon; this is referred to as CRISPR-stop. Its application in generating zygotic mutations has not been well explored yet. Here, we first performed a proof-of-principle test by disrupting Atoh1, a gene crucial for auditory hair cell generation. Next, we individually mutated vGlut3 (Slc17a8), otoferlin (Otof) and prestin (Slc26a5), three genes needed for normal hearing function. Finally, we successfully disrupted vGlut3, Otof and prestin simultaneously. Our results show that CRISPR-stop can efficiently generate single or triple homozygous F0 mouse mutants, bypassing laborious mouse breeding. We believe that CRISPR-stop is a powerful method that will pave the way for high-throughput screening of mouse developmental and functional genes, matching the efficiency of methods available for model organisms such as Drosophila.

Whole-genome sequencing of Chinese centenarians reveals important genetic variants in aging WGS of centenarian for genetic analysis of aging.

BACKGROUND: Genetic research on longevity has provided important insights into the mechanism of aging and aging-related diseases. Pinpointing import genetic variants associated with aging could provide insights for aging research. METHODS: We performed a whole-genome sequencing in 19 centenarians to establish the genetic basis of human longevity. RESULTS: Using SKAT analysis, we found 41 significantly correlated genes in centenarians as compared to control genomes. Pathway enrichment analysis of these genes showed that immune-related pathways were enriched, suggesting that immune pathways might be critically involved in aging. HLA typing was next performed based on the whole-genome sequencing data obtained. We discovered that several HLA subtypes were significantly overrepresented. CONCLUSIONS: Our study indicated a new mechanism of longevity, suggesting potential genetic variants for further study.

In vivo ectopic Ngn1 and Neurod1 convert neonatal cochlear glial cells into spiral ganglion neurons.

Damage or degeneration of inner ear spiral ganglion neurons (SGNs) causes hearing impairment. Previous in vitro studies indicate that cochlear glial cells can be reprogrammed into SGNs, however, it remains unknown whether this can occur in vivo. Here, we show that neonatal glial cells can be converted, in vivo, into SGNs (defined as new SGNs) by simultaneous induction of Neurog1 (Ngn1) and Neurod1. New SGNs express SGN markers, Tuj1, Map2, Prox1, Mafb and Gata3, and reduce glial cell marker Sox10 and Scn7a. The heterogeneity within new SGNs is illustrated by immunostaining and transcriptomic assays. Transcriptomes analysis indicates that well reprogrammed SGNs are similar to type I SGNs. In addition, reprogramming efficiency is positively correlated with the dosage of Ngn1 and Neurod1, but declined with aging. Taken together, our in vivo data demonstrates the plasticity of cochlear neonatal glial cells and the capacity of Ngn1 and Neurod1 to reprogram glial cells into SGNs. Looking ahead, we expect that combination of Neurog1 and Neurod1 along with other factors will further boost the percentage of fully converted (Mafb+/Gata3+) new SGNs.

Application and prospect of cutting-edge information technology in biomedical big data.

In recent years, with the development of various high-throughput omics based biological technologies (BT), biomedical research began to enter the era of big data. In the face of high-dimensional, multi-domain and multi-modal biomedical big data, scientific research requires a new paradigm of data intensive scientific research. The vigorous development of cutting-edge information technologies (IT) such as cloud computing, blockchain and artificial intelligence provides technical means for the practice of this new research paradigm. Here,we describe the application of such cutting-edge information technologies in biomedical big data, and propose a forward-looking prospect for the construction of a new paradigm supporting environment for data intensive scientific research. We expect to establish a new research scheme and new scientific research paradigm integrating BT & IT technology, which can finally promote the great leap forward development of biomedical research.

A novel virtual barcode strategy for accurate panel-wide variant calling in circulating tumor DNA.

BACKGROUND: Hybrid capture-based next-generation sequencing of DNA has been widely applied in the detection of circulating tumor DNA (ctDNA). Various methods have been proposed for ctDNA detection, but low-allelic-fraction (AF) variants are still a great challenge. In addition, no panel-wide calling algorithm is available, which hiders the full usage of ctDNA based 'liquid biopsy'. Thus, we developed the VBCALAVD (Virtual Barcode-based Calling Algorithm for Low Allelic Variant Detection) in silico to overcome these limitations. RESULTS: Based on the understanding of the nature of ctDNA fragmentation, a novel platform-independent virtual barcode strategy was established to eliminate random sequencing errors by clustering sequencing reads into virtual families. Stereotypical mutant-family-level background artifacts were polished by constructing AF distributions. Three additional robust fine-tuning filters were obtained to eliminate stochastic mutant-family-level noises. The performance of our algorithm was validated using cell-free DNA reference standard samples (cfDNA RSDs) and normal healthy cfDNA samples (cfDNA controls). For the RSDs with AFs of 0.1, 0.2, 0.5, 1 and 5%, the mean F1 scores were 0.43 (0.25~0.56), 0.77, 0.92, 0.926 (0.86~1.0) and 0.89 (0.75~1.0), respectively, which indicates that the proposed approach significantly outperforms the published algorithms. Among controls, no false positives were detected. Meanwhile, characteristics of mutant-family-level noise and quantitative determinants of divergence between mutant-family-level noises from controls and RSDs were clearly depicted. CONCLUSIONS: Due to its good performance in the detection of low-AF variants, our algorithm will greatly facilitate the noninvasive panel-wide detection of ctDNA in research and clinical settings. The whole pipeline is available at https://github.com/zhaodalv/VBCALAVD.

Establishment of a hepatocellular carcinoma patient-derived xenograft platform and its application in biomarker identification.

Using a method optimized in hepatocellular carcinoma (HCC), we established patient-derived xenograft (PDX) models with an increased take rate (42.2%) and demonstrated that FBS +10% dimethyl sulfoxide exhibited the highest tumor take rate efficacy. Among 254 HCC patients, 103 stably transplantable xenograft lines that could be serially passaged, cryopreserved and revived were established. These lines maintained the diversity of HCC and the essential features of the original specimens at the histological, transcriptome, proteomic and genomic levels. Tumor engraftment was associated with lack of encapsulation, poor tumor differentiation, large size and overexpression of cancer stem cell biomarkers, and was an independent predictor for overall survival and tumor recurrence after resection. To confirm the preclinical value of the PDX model in HCC treatment, several antitumor agents were tested in 16 selected PDX models. The results revealed a high degree of pharmacologic heterogeneity in the cohort, as well as heterogeneity to different agents in the same individual. The sorafenib responses observed between HCC patients and the corresponding PDXs were also consistent. After molecular characterization of the PDX models, we explored the predictive markers for sorafenib response and found that mitogen-activated protein kinase kinase kinase 1 (MAP3K1) might play an important role in sorafenib resistance and sorafenib response is impaired in patients with MAP3K1 downexpression. Our results indicated that PDX models could accurately reproduce patient tumors biology and could aid in the discovery of new treatments to advance in precision medicine.

Combination of multiple tumor-infiltrating immune cells predicts clinical outcome in colon cancer.

The infiltration of immune cells is highly associated with the development and progression of cancer. Thus, integrating the immune cell infiltrating profile into an immune cell infiltrating score may predict the survival of cancer patients. Here, by combining the infiltration proportion of 22 immune cells inferred from bulk tumor transcriptome of 879 patients, we identified an immune cell infiltrating indicator including five types of immune cells: resting T cells CD4 memory, macrophages M0-M2, and activated mast cells. The signature distinguished patients into two groups (high-risk and low-risk) with significantly different survival in the training cohort (HR = 1.96, 95% CI = 1.29-2.98, P = .0013) and two additional cohorts (HR = 1.78, 95%, CI = 1.16-2.75, P = .0079 and HR = 2.01, 95% CI = 1.28-3.14, P = .0019). The indicator remained as an independent prognostic factor after adjusting for clinicopathological factors by multivariable analysis in all cohorts. Stratification analysis showed that the signature consistently and significantly predicted survival of high-stage colon cancer patients in the training cohort (P = .00053) and validation cohorts (P = .017 and P = .0035). Moreover, we found that the low-risk patients were significantly correlated with deficient mismatch repair and the high-risk patients had a weak ability of trafficking of immune cells to tumors in the cancer immunity cycle. Overall, our results showed that integrating multiple tumor-infiltrating immune cells was an effective strategy for uncovering robust prognostic factor for tumor patients, and potentially was a promising response marker for precision oncology to be explored.

The self-organization model reveals systematic characteristics of aging.

BACKGROUND: Aging is a fundamental biological process, where key bio-markers interact with each other and synergistically regulate the aging process. Thus aging dysfunction will induce many disorders. Finding aging markers and re-constructing networks based on multi-omics data (i.e. methylation, transcriptional and so on) are informative to study the aging process. However, optimizing the model to predict aging have not been performed systemically, although it is critical to identify potential molecular mechanism of aging related diseases. METHODS: This paper aims to model the aging self-organization system using a series of supervised learning methods, and study complex molecular mechanisms of aging at system level: i.e. optimizing the aging network; summarizing interactions between aging markers; accumulating patterns of aging markers within module; finding order-parameters in the aging self-organization system. RESULTS: In this work, the normal aging process is modeled based on multi-omics profiles across different tissues. In addition, the computational pipeline aims to model aging self-organizing systems and study the relationship between aging and related diseases (i.e. cancers), thus provide useful indicators of aging related diseases and could help to improve prediction abilities of diagnostics. CONCLUSIONS: The aging process could be studied thoroughly by modelling the self-organization system, where key functions and the crosstalk between aging and cancers were identified.