Search

Motivation Novel machine learning and statistical modeling studies rely on standardized comparisons to existing methods using well-studied benchmark datasets. Few tools exist that provide rapid access to many of these datasets through a standardized, user-friendly interface that integrates well with popular data science workflows. Results This release of PMLB (Penn Machine Learning Benchmarks) provides the largest collection of diverse, public benchmark datasets for evaluating new machine learning and data science methods aggregated in one location. v1.0 introduces a number of critical improvements developed following discussions with the open-source community. Availability and implementation PMLB is available at https://github.com/EpistasisLab/pmlb. Python and R interfaces for PMLB can be installed through the Python Package Index and Comprehensive R Archive Network, respectively.

Automated machine learning (AutoML) and artificial neural networks (ANNs) have revolutionized the field of artificial intelligence by yielding incredibly high-performing models to solve a myriad of inductive learning tasks. In spite of their successes, little guidance exists on when to use one versus the other. Furthermore, relatively few tools exist that allow the integration of both AutoML and ANNs in the same analysis to yield results combining both of their strengths. Here, we present TPOT-NN—a new extension to the tree-based AutoML software TPOT—and use it to explore the behavior of automated machine learning augmented with neural network estimators (AutoML+NN), particularly when compared to non-NN AutoML in the context of simple binary classification on a number of public benchmark datasets. Our observations suggest that TPOT-NN is an effective tool that achieves greater classification accuracy than standard tree-based AutoML on some datasets, with no loss in accuracy on others. We also provide preliminary guidelines for performing AutoML+NN analyses, and recommend possible future directions for AutoML+NN methods research, especially in the context of TPOT.

Background Perfluorooctanoic acid (PFOA), is a persistent fluorinated compound with oil and water repelling properties found in cookware, food packaging and municipal water systems. Adult animals exposed to PFOA develop hepatomegaly, fatty liver, peroxisome proliferation, and immunotoxicity. Rodents exposed to PFOA in utero have altered hepatic lipid metabolism, increased hepatic de novo lipogenesis and susceptibility to non-alcoholic fatty liver disease (NAFLD), but underlying molecular mechanisms remain unknown. With increasing rates of obesity, diabetes, and NAFLD it is critical to examine the mechanisms by which in utero exposure to PFOA contributes to the development of metabolic syndrome in offspring. Objectives To characterize how PFOA exposure during hepatocyte differentiation leads to the development of NAFLD through alterations in DNA methylation profiles and changes in the availability of transcription factor binding sites. Methods HepaRG cells (human-derived hepatocyte progenitor cells) were treated with 0.5uM PFOA or vehicle for 48 hours followed by differentiation. Undifferentiated and differentiated hepatocytes exposed to PFOA were assessed relative to controls (n=4). RNASeq was completed; DESeq2 identified differentially expressed genes via false discovery rate (FDR) of 10%, and FDR of Results PFOA treatment resulted in decreased expression of the transcription factors early growth response protein 1 (EGR1), nuclear receptor Nur77 (NR4A1), early growth response protein 2 (EGR2), krueppel-like factor 10 (KLF10) and fos-related antigen 1 (FOSL1), which are key genes linked to impaired hepatic insulin signaling, lipid metabolism, steatosis and fibrosis (q Conclusions We conclude hepatocyte progenitor cells exposed to low dose PFOA results in changes in DNA methylation and expression of key metabolic genes linked to NAFLD, notably EGR1, a gene previously linked to NAFLD, suggesting PFOA exposure in utero has lasting effects. Presentation: Sunday, June 12, 2022 12:30 p.m. - 2:30 p.m., Sunday, June 12, 2022 12:48 p.m. - 12:53 p.m.

Scale drop disease virus (SDDV), an emerging piscine iridovirus prevalent in farmed Asian seabass Lates calcarifer in Southeast Asia, was firstly scientifically descripted in Singapore in 2015. Here, an SDDV isolate ZH-06/20 was isolated by inoculating filtered ascites from diseased juvenile yellowfin seabream into MFF-1 cell. Advanced cytopathic effects were observed 6 days post-inoculation. A transmission electron microscopy examination confirmed that numerous virion particles, about 140 nm in diameter, were observed in infected MFF-1 cell. ZH-06/20 was further purified and both whole genome and virion proteome were determined. The results showed that ZH-06/20 was composed of 131,122 bp with 135 putative viral proteins and 113 of them were further detected by virion proteome. Western blot analysis showed that no (or weak) cross-reaction was observed among several major viral proteins between ZH-06/20 and ISKNV-like megalocytivirus. An artificial challenge showed that ZH-06/20 could cause 100% death to juvenile yellowfin seabream. A typical sign was characterized by severe ascites, but not scale drop, which was considerably different from SDD syndrome in Asian seabass. Collectively, SDDV was confirmed, for the first time, as the causative agent of ascites diseases in farmed yellowfin seabream. Our study offers useful information to better understanding SDDV-associated diseases in farmed fish.

The impact of intratumoral heterogeneity (ITH) and the resultant neoantigen landscape on T cell immunity are poorly understood. ITH is a widely recognized feature of solid tumors and poses distinct challenges related to the development of effective therapeutic strategies, including cancer neoantigen vaccines. Here, we performed deep targeted DNA sequencing of multiple metastases from melanoma patients and observed ubiquitous sharing of clonal and subclonal single nucleotide variants (SNVs) encoding putative HLA class I-restricted neoantigen epitopes. However, spontaneous antitumor CD8+ T cell immunity in peripheral blood and tumors was restricted to a few clonal neoantigens featuring an oligo-/monoclonal T cell-receptor (TCR) repertoire. Moreover, in various tumors of the 4 patients examined, no neoantigen-specific TCR clonotypes were identified despite clonal neoantigen expression. Mature dendritic cell (mDC) vaccination with tumor-encoded amino acid-substituted (AAS) peptides revealed diverse neoantigen-specific CD8+ T responses, each composed of multiple TCR clonotypes. Isolation of T cell clones by limiting dilution from tumor-infiltrating lymphocytes (TILs) permitted functional validation regarding neoantigen specificity. Gene transfer of TCRαβ heterodimers specific for clonal neoantigens confirmed correct TCR clonotype assignments based on high-throughput TCRBV CDR3 sequencing. Our findings implicate immunological ignorance of clonal neoantigens as the basis for ineffective T cell immunity to melanoma and support the concept that therapeutic vaccination, as an adjunct to checkpoint inhibitor treatment, is required to increase the breadth and diversity of neoantigen-specific CD8+ T cells.

The c-Myc oncogene drives malignant progression and induces robust anabolic and proliferative programmes leading to intrinsic stress. The mechanisms enabling adaptation to MYC-induced stress are not fully understood. Here we reveal an essential role for activating transcription factor 4 (ATF4) in survival following MYC activation. MYC upregulates ATF4 by activating general control nonderepressible 2 (GCN2) kinase through uncharged transfer RNAs. Subsequently, ATF4 co-occupies promoter regions of over 30 MYC-target genes, primarily those regulating amino acid and protein synthesis, including eukaryotic translation initiation factor 4E-binding protein 1 (4E-BP1), a negative regulator of translation. 4E-BP1 relieves MYC-induced proteotoxic stress and is essential to balance protein synthesis. 4E-BP1 activity is negatively regulated by mammalian target of rapamycin complex 1 (mTORC1)-dependent phosphorylation and inhibition of mTORC1 signalling rescues ATF4-deficient cells from MYC-induced endoplasmic reticulum stress. Acute deletion of ATF4 significantly delays MYC-driven tumour progression and increases survival in mouse models. Our results establish ATF4 as a cellular rheostat of MYC activity, which ensures that enhanced translation rates are compatible with survival and tumour progression.

Machine Learning (ML) approaches are increasingly being used in biomedical applications. Important challenges of ML include choosing the right algorithm and tuning the parameters for optimal performance. Automated ML (AutoML) methods, such as Tree-based Pipeline Optimization Tool (TPOT), have been developed to take some of the guesswork out of ML thus making this technology available to users from more diverse backgrounds. The goals of this study were to assess applicability of TPOT to genomics and to identify combinations of single nucleotide polymorphisms (SNPs) associated with coronary artery disease (CAD), with a focus on genes with high likelihood of being good CAD drug targets. We leveraged public functional genomic resources to group SNPs into biologically meaningful sets to be selected by TPOT. We applied this strategy to data from the UK Biobank, detecting a strikingly recurrent signal stemming from a group of 28 SNPs. Importance analysis of these uncovered functional relevance of the top SNPs to genes whose association with CAD is supported in the literature and other resources. Furthermore, we employed game-theory based metrics to study SNP contributions to individual level TPOT predictions and discover distinct clusters of well-predicted CAD cases. The latter indicates a promising approach towards precision medicine.

Fibroblast growth factor receptors (FGFRs) are a family of receptor tyrosine kinases expressed on the cell membrane that play crucial roles in cellular lineage commitment, differentiation, proliferation, and apoptosis. Deregulated FGFR signaling is observed in a subset of tumors across various histologies, making FGFRs ideal therapeutic targets. We sought to determine the genetic landscape of FGFR-family variations in a cohort of pediatric and young adult patients with solid tumors. The CHOP Comprehensive Solid Tumor Panel was performed on 1,420 patients. The panel covers 238 cancer genes and screens for single nucleotide variants (SNVs), indels, copy number alterations, and 117 fusion gene partners interrogating over 700 exons for known and novel fusions. Identified variants were categorized and reported according to the AMP/ASCO/CAP guidelines. Fifty-six patients (4.1%), including 47 children and 9 young adults, were found to carry at least one FGFR alteration in their tumors. CNS tumors accounted for most of the cases (51 total, 87.9%), with pilomyxoid astrocytoma/pilocytic astrocytoma and dysembryoplastic neuroepithelial tumor the most common (13 and 12 patients, respectively). Non-CNS solid tumors included rhabdomyosarcoma (4 patients), neuroblastoma/ganglioneuroblastoma (2), and follicular thyroid carcinoma (1). FGFR somatic alterations were found in 56 tumors including 41 SNVs and small indels, 6 internal tandem duplications (ITDs), and 15 fusions genes. The most common SNVs observed were hotspot mutations p.K656E and p.N546K of FGFR1. Sequence alterations in FGFR1 contained 35 SNVs and small indels, mostly gain of function mutations located in the kinase domain, and 6 kinase domain ITDs. One SNV was identified in FGFR2 in the immunoglobulin domain. Two SNVs were reported in FGFR3, both of which were in the fibroblast growth factor receptor family domain, and 3 SNVs were identified in FGFR4, all occurring at the p.V550 codon located on the kinase domain. Companion mutations in non-FGFR genes were detected in 27 tumors, predominantly involving RAS signaling pathway genes including NF1 (14 variants), PIK3CA (8), PTPN11 (6) and PIK3R1 (4). Among fusion variants, FGFR1-TACC1 fusions were found in 5 patients, mostly in pediatric patients. One FGFR3-TACC3 fusion was identified in one young adult patient. Seven pediatric patients tested positive for FGFR2 fusions; all with different 3’ partners. The detection of an FGFR alteration defined or changed the histologic diagnosis for 22 patients. Our results reveal that FGFR alterations account for 4.1% (56/1420) of the patients with solid tumors tested in our laboratory. The majority of the FGFR-positive tumors are low-grade CNS tumors. Further, the identification of FGFR alterations can significantly improve the tumor diagnosis and provide genomic evidence for potential targeted treatment with FGFR inhibitors. Citation Format: Jinhua Wu, Jeffrey Schubert, Feng Xu, Ariel Long, Maha Patel, Netta Golenberg, Weixuan Fu, Kajia Cao, Jiani Chen, Elizabeth H. Denenberg, Elizabeth A. Fanning, Rochelle Bagatell, Theodore W. Laetsch, Adam Resnick, Mariarita Santi, Phillip Jay B. Storm, Minjie Luo, Lea F. Surrey, Yiming Zhong, Marilyn M. Li. The spectrum of FGFR mutations in pediatric and young adult solid tumor [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 5268.

Many approaches to identify therapeutically relevant neoantigens couple tumor sequencing with bioinformatic algorithms and inferred rules of tumor epitope immunogenicity. However, there are no reference data to compare these approaches, and the parameters governing tumor epitope immunogenicity remain unclear. Here, we assembled a global consortium wherein each participant predicted immunogenic epitopes from shared tumor sequencing data. 608 epitopes were subsequently assessed for T cell binding in patient-matched samples. By integrating peptide features associated with presentation and recognition, we developed a model of tumor epitope immunogenicity that filtered out 98% of non-immunogenic peptides with a precision above 0.70. Pipelines prioritizing model features had superior performance, and pipeline alterations leveraging them improved prediction performance. These findings were validated in an independent cohort of 310 epitopes prioritized from tumor sequencing data and assessed for T cell binding. This data resource enables identification of parameters underlying effective anti-tumor immunity and is available to the research community.

Background A typical task in bioinformatics consists of identifying which features are associated with a target outcome of interest and building a predictive model. Automated machine learning (AutoML) systems such as the Tree-based Pipeline Optimization Tool (TPOT) constitute an appealing approach to this end. However, in biomedical data, there are often baseline characteristics of the subjects in a study or batch effects that need to be adjusted for in order to better isolate the effects of the features of interest on the target. Thus, the ability to perform covariate adjustments becomes particularly important for applications of AutoML to biomedical big data analysis. Results We developed an approach to adjust for covariates affecting features and/or target in TPOT. Our approach is based on regressing out the covariates in a manner that avoids ‘leakage’ during the cross-validation training procedure. We describe applications of this approach to toxicogenomics and schizophrenia gene expression data sets. The TPOT extensions discussed in this work are available at https://github.com/EpistasisLab/tpot/tree/v0.11.1-resAdj. Conclusions In this work, we address an important need in the context of AutoML, which is particularly crucial for applications to bioinformatics and medical informatics, namely covariate adjustments. To this end we present a substantial extension of TPOT, a genetic programming based AutoML approach. We show the utility of this extension by applications to large toxicogenomics and differential gene expression data. The method is generally applicable in many other scenarios from the biomedical field.

Additional file 5. resAdj TPOT pre-processor output. Structure and column description of the output from this pre-processor.

Additional file 4. Additional methods. Details for the leakage-free covariate adjustment.

Background: Gestational diabetes (GDM) has profound effects on the intrauterine metabolic milieu, induces marked abnormalities in fetal glucose and insulin secretion and is linked to obesity and diabetes in the offspring, but the mechanisms remain largely unknown. Epigenetic modifications in stems cells may be one mechanism by which an in utero exposure can lead to the development of diabetes and obesity later in life. Objective: To identify novel pathways contributing to the development of diabetes and obesity in offspring exposed to GDM in utero by integrating data generated from transcriptome and methylome analysis from second trimester human amniocytes exposed to GDM in utero. Methods: We analyzed RNAseq and genome wide DNA methylation data (ERRBS) generated from second trimester amniocytes obtained from women with GDM (n=14). Amniocytes have stem cells-like characteristics and are derived from the fetus. Expression data of 22,271 genes were retrieved from RNAseq data. CpGs with significant changes in DNA methylation were mapped into 20,028 genes by collapsing methylation probes into promoter and gene regions. To better understand the associations among diverse gene sets or among gene sets and GDM,we first constructed two weighted co-expression networks from RNAseq and DNA methylation data, respectively. Then, two co-expression networks were integrated using a linear combination. With the integrated co-expression network, graph-based label propagation algorithm was employed to prioritize GDM-associated genes. Results: From the differential expression analysis between GDM and control, the top 20 query genes, including 11 genes and 9 methylated genes, were selected for label propagation. Finally, the top 100 genes were picked up for the pathway enrichment testusing an over-representation analysis approach. Significantly enriched pathways included: Interferon Signaling, N-glycan Antennae Elongation, Sphingolipid Pathway and Metabolism, Classical Complement Pathway, Complement and Coagulation Cascades, Tryptophan Metabolism, Peroxisomal Protein Import, Unsaturated Fatty Acid Metabolism, Complement Activation, Human Innate Immune Response, Ceramide Metabolism, Fertilization Pathway, Keratan Sulfate Biosynthesis Pathway, Transport to the Golgi and Modification Pathway (FDR q Conclusion: Using a novel bioinformatic approach to synthesize transcriptome and methylome data derived from human amniocytes exposed to GDM in utero, we identified potential pathways involved in programming of diabetes and obesity in offspring including pathways involving the immune response, complex lipid metabolism, the complement pathway, and protein transport and processing. Further investigation of these pathways may yield new mechanisms contributing to diabetes and obesity.

Background: Perfluorooctanoic acid (PFOA), is a persistent fluorinated compound with oil and water repelling properties found in cookware, food packaging and municipal water systems. Adult animals exposed to PFOA develop hepatomegaly, fatty liver, peroxisome proliferation, and immunotoxicity. Rodents exposed to PFCs in utero have altered hepatic lipid metabolism, increased hepatic de novo lipogenesis and susceptibility to non-alcoholic fatty liver disease (NAFLD), but underlying molecular mechanisms remain unknown. With increasing rates of obesity, diabetes, and NAFLD it is critical to examine the mechanisms by which in utero exposure to PFOA contributes to the development of metabolic syndrome in offspring. Objective: To determine mechanism by which PFOA alters gene expression in undifferentiated hepatic progenitor cells. Design/methods: HepaRG cells, a human derived hepatocyte progenitor cell line, was treated with 0.5uM PFOA or vehicle for 48 hours followed by differentiation into hepatocytes. Total RNA was extracted using the RNeasy (Qiagen) [total RNA A260/280>2 and RNA integrity number >7 (Agilent Bioanalyzer)] to generate libraries with the Illumina TruSeq stranded total RNA kit. RNA-Seq was performed using 85 bp single-end read sequencing to generate >20 million reads per sample. RNAseq data was aligned to hg38 using STAR v2.6.1a and then quantified with featureCounts v1.6.2. DESeq2 identified differentially expressed genes via FDR (false discovery rate) after Bonferroni correction. Differentially expressed gene lists were used for Ingenuity Pathway Analysis (IPA) to identify pathways of biological significance. Results: PFOA treatment resulted in increased expression of transcription factors EGR1 (early growth response protein 1), NR4A1 (nuclear receptor Nur77), EGR2 (early growth response protein 2), KLF10 (Krueppel-like factor 10) and FOSL1 (Fos-related antigen 1), key genes linked to impaired hepatic insulin signaling, hepatic lipid metabolism, steatosis and fibrosis (fold change > 1.5; q

Evolutionary computation (EC) has been widely applied to biological and biomedical data. The practice of EC involves the tuning of many parameters, such as population size, generation count, selection size, and crossover and mutation rates. Through an

Fish nocardiosis is a chronic granulomatous bacterial disease mainly caused by three pathogenic bacteria, including Nocardia seriolae, N. asteroids and N. salmonicida. Molecular chaperone DnaK and GroEL were identified to be the common antigens of the three pathogenic Nocardia species in our previous studies. To evaluate the immune protective effect of two DNA vaccines encoding DnaK or GroEL against fish nocardiosis, hybrid snakehead were vaccinated and the immune responses induced by these two vaccines were comparatively analyzed. The results suggested it needed at least 7 d to transport DnaK or GroEL gene from injected muscle to head kidney, spleen and liver and stimulate host's immune system for later protection after immunization by DNA vaccines. Additionally, non-specific immunity parameters (serum lysozyme (LYZ), peroxidase (POD), acid phosphatase (ACP), alkaline phosphatase (AKP) and superoxide dismutase (SOD) activities), specific antibody (IgM) production and immune-related genes (MHCIα, MHCIIα, CD4, CD8α, IL-1β and TNFα) were used to evaluate the immune responses induced in vaccinated hybrid snakehead. It proved that all the above-mentioned immune activities were significantly enhanced after immunization with these two DNA vaccines. The protective efficacy of pcDNA-DnaK and pcDNA-GroEL DNA vaccines, in terms of relative percentage survival (RPS), were 53.01% and 80.71% respectively. It demonstrated that these two DNA vaccines could increase the survival rate of hybrid snakehead against fish nocardiosis, albeit with variations in immunoprotective effects. Taken together, these results indicated that both pcDNA-DnaK and pcDNA-GroEL DNA vaccines could boost the innate, humoral and cellular immune response in hybrid snakehead and show highly protective efficacy against fish nocardiosis, suggesting that DnaK and GroEL were promising vaccine candidates. These findings will promote the development of DNA vaccines against fish nocardiosis in aquaculture.

Motivation: Many researchers with domain expertise are unable to easily apply machine learning to their bioinformatics data due to a lack of machine learning and/or coding expertise. Methods that have been proposed thus far to automate machine learning mostly require programming experience as well as expert knowledge to tune and apply the algorithms correctly. Here, we study a method of automating biomedical data science using a web-based platform that uses AI to recommend model choices and conduct experiments. We have two goals in mind: first, to make it easy to construct sophisticated models of biomedical processes; and second, to provide a fully automated AI agent that can choose and conduct promising experiments for the user, based on the user's experiments as well as prior knowledge. To validate this framework, we experiment with hundreds of classification problems, comparing to state-of-the-art, automated approaches. Finally, we use this tool to develop predictive models of septic shock in critical care patients. Results: We find that matrix factorization-based recommendation systems outperform meta-learning methods for automating machine learning. This result mirrors the results of earlier recommender systems research in other domains. The proposed AI is competitive with state-of-the-art automated machine learning methods in terms of choosing optimal algorithm configurations for datasets. In our application to prediction of septic shock, the AI-driven analysis produces a competent machine learning model (AUROC 0.85 +/- 0.02) that performs on par with state-of-the-art deep learning results for this task, with much less computational effort., Comment: 17 pages, 8 figures. this version fixes link to pennai in abstract

Through an extensive series of experiments over multiple evolutionary algorithm implementations and 25 problems we showed that parameter space tends to be rife with viable parameters, somewhat in contrast with common lore [6].

While artificial intelligence (AI) has become widespread, many commercial AI systems are not yet accessible to individual researchers nor the general public due to the deep knowledge of the systems required to use them. We believe that AI has matured to the point where it should be an accessible technology for everyone. We present an ongoing project whose ultimate goal is to deliver an open source, user-friendly AI system that is specialized for machine learning analysis of complex data in the biomedical and health care domains. We discuss how genetic programming can aid in this endeavor, and highlight specific examples where genetic programming has automated machine learning analyses in previous projects.