Global and local functions of the Fused kinase ortholog CdaH in intracellular patterning in Tetrahymena.

Ciliates assemble numerous microtubular structures into complex cortical patterns. During ciliate division, the pattern is duplicated by intracellular segmentation that produces a tandem of daughter cells. In Tetrahymena thermophila, the induction and positioning of the division boundary involves two mutually antagonistic factors: posterior CdaA (cyclin E) and anterior CdaI (Hippo kinase). Here, we characterized the related cdaH-1 allele, which confers a pleiotropic patterning phenotype including an absence of the division boundary and an anterior-posterior mispositioning of the new oral apparatus. CdaH is a Fused or Stk36 kinase ortholog that localizes to multiple sites that correlate with the effects of its loss, including the division boundary and the new oral apparatus. CdaH acts downstream of CdaA to induce the division boundary and drives asymmetric cytokinesis at the tip of the posterior daughter. CdaH both maintains the anterior-posterior position of the new oral apparatus and interacts with CdaI to pattern ciliary rows within the oral apparatus. Thus, CdaH acts at multiple scales, from induction and positioning of structures on the cell-wide polarity axis to local organelle-level patterning.

A novel membrane complex is required for docking and regulated exocytosis of lysosome-related organelles in Tetrahymena thermophila.

In the ciliate Tetrahymena thermophila, lysosome-related organelles called mucocysts accumulate at the cell periphery where they secrete their contents in response to extracellular events, a phenomenon called regulated exocytosis. The molecular bases underlying regulated exocytosis have been extensively described in animals but it is not clear whether similar mechanisms exist in ciliates or their sister lineage, the Apicomplexan parasites, which together belong to the ecologically and medically important superphylum Alveolata. Beginning with a T. thermophila mutant in mucocyst exocytosis, we used a forward genetic approach to uncover MDL1 (Mucocyst Discharge with a LamG domain), a novel gene that is essential for regulated exocytosis of mucocysts. Mdl1p is a 40 kDa membrane glycoprotein that localizes to mucocysts, and specifically to a tip domain that contacts the plasma membrane when the mucocyst is docked. This sub-localization of Mdl1p, which occurs prior to docking, underscores a functional asymmetry in mucocysts that is strikingly similar to that of highly polarized secretory organelles in other Alveolates. A mis-sense mutation in the LamG domain results in mucocysts that dock but only undergo inefficient exocytosis. In contrast, complete knockout of MDL1 largely prevents mucocyst docking itself. Mdl1p is physically associated with 9 other proteins, all of them novel and largely restricted to Alveolates, and sedimentation analysis supports the idea that they form a large complex. Analysis of three other members of this putative complex, called MDD (for Mucocyst Docking and Discharge), shows that they also localize to mucocysts. Negative staining of purified MDD complexes revealed distinct particles with a central channel. Our results uncover a novel macromolecular complex whose subunits are conserved within alveolates but not in other lineages, that is essential for regulated exocytosis in T. thermophila.

AggMapNet: enhanced and explainable low-sample omics deep learning with feature-aggregated multi-channel networks.

Omics-based biomedical learning frequently relies on data of high-dimensions (up to thousands) and low-sample sizes (dozens to hundreds), which challenges efficient deep learning (DL) algorithms, particularly for low-sample omics investigations. Here, an unsupervised novel feature aggregation tool AggMap was developed to Aggregate and Map omics features into multi-channel 2D spatial-correlated image-like feature maps (Fmaps) based on their intrinsic correlations. AggMap exhibits strong feature reconstruction capabilities on a randomized benchmark dataset, outperforming existing methods. With AggMap multi-channel Fmaps as inputs, newly-developed multi-channel DL AggMapNet models outperformed the state-of-the-art machine learning models on 18 low-sample omics benchmark tasks. AggMapNet exhibited better robustness in learning noisy data and disease classification. The AggMapNet explainable module Simply-explainer identified key metabolites and proteins for COVID-19 detections and severity predictions. The unsupervised AggMap algorithm of good feature restructuring abilities combined with supervised explainable AggMapNet architecture establish a pipeline for enhanced learning and interpretability of low-sample omics data.

Clinical trials, progression-speed differentiating features and swiftness rule of the innovative targets of first-in-class drugs.

Drugs produce their therapeutic effects by modulating specific targets, and there are 89 innovative targets of first-in-class drugs approved in 2004-17, each with information about drug clinical trial dated back to 1984. Analysis of the clinical trial timelines of these targets may reveal the trial-speed differentiating features for facilitating target assessment. Here we present a comprehensive analysis of all these 89 targets, following the earlier studies for prospective prediction of clinical success of the targets of clinical trial drugs. Our analysis confirmed the literature-reported common druggability characteristics for clinical success of these innovative targets, exposed trial-speed differentiating features associated to the on-target and off-target collateral effects in humans and further revealed a simple rule for identifying the speedy human targets through clinical trials (from the earliest phase I to the 1st drug approval within 8 years). This simple rule correctly identified 75.0% of the 28 speedy human targets and only unexpectedly misclassified 13.2% of 53 non-speedy human targets. Certain extraordinary circumstances were also discovered to likely contribute to the misclassification of some human targets by this simple rule. Investigation and knowledge of trial-speed differentiating features enable prioritized drug discovery and development.

LF4/MOK and a CDK-related kinase regulate the number and length of cilia in Tetrahymena.

The length of cilia is controlled by a poorly understood mechanism that involves members of the conserved RCK kinase group, and among them, the LF4/MOK kinases. The multiciliated protist model, Tetrahymena, carries two types of cilia (oral and locomotory) and the length of the locomotory cilia is dependent on their position with the cell. In Tetrahymena, loss of an LF4/MOK ortholog, LF4A, lengthened the locomotory cilia, but also reduced their number. Without LF4A, cilia assembled faster and showed signs of increased intraflagellar transport (IFT). Consistently, overproduced LF4A shortened cilia and downregulated IFT. GFP-tagged LF4A, expressed in the native locus and imaged by total internal reflection microscopy, was enriched at the basal bodies and distributed along the shafts of cilia. Within cilia, most LF4A-GFP particles were immobile and a few either diffused or moved by IFT. We suggest that the distribution of LF4/MOK along the cilium delivers a uniform dose of inhibition to IFT trains that travel from the base to the tip. In a longer cilium, the IFT machinery may experience a higher cumulative dose of inhibition by LF4/MOK. Thus, LF4/MOK activity could be a readout of cilium length that helps to balance the rate of IFT-driven assembly with the rate of disassembly at steady state. We used a forward genetic screen to identify a CDK-related kinase, CDKR1, whose loss-of-function suppressed the shortening of cilia caused by overexpression of LF4A, by reducing its kinase activity. Loss of CDKR1 alone lengthened both the locomotory and oral cilia. CDKR1 resembles other known ciliary CDK-related kinases: LF2 of Chlamydomonas, mammalian CCRK and DYF-18 of C. elegans, in lacking the cyclin-binding motif and acting upstream of RCKs. The new genetic tools we developed here for Tetrahymena have potential for further dissection of the principles of cilia length regulation in multiciliated cells.

CMAUP: a database of collective molecular activities of useful plants.

The beneficial effects of functionally useful plants (e.g. medicinal and food plants) arise from the multi-target activities of multiple ingredients of these plants. The knowledge of the collective molecular activities of these plants facilitates mechanistic studies and expanded applications. A number of databases provide information about the effects and targets of various plants and ingredients. More comprehensive information is needed for broader classes of plants and for the landscapes of individual plant's multiple targets, collective activities and regulated biological pathways, processes and diseases. We therefore developed a new database, Collective Molecular Activities of Useful Plants (CMAUP), to provide the collective landscapes of multiple targets (ChEMBL target classes) and activity levels (in 2D target-ingredient heatmap), and regulated gene ontologies (GO categories), biological pathways (KEGG categories) and diseases (ICD blocks) for 5645 plants (2567 medicinal, 170 food, 1567 edible, 3 agricultural and 119 garden plants) collected from or traditionally used in 153 countries and regions. These landscapes were derived from 47 645 plant ingredients active against 646 targets in 234 KEGG pathways associated with 2473 gene ontologies and 656 diseases. CMAUP (http://bidd2.nus.edu.sg/CMAUP/) is freely accessible and searchable by keywords, plant usage classes, species families, targets, KEGG pathways, gene ontologies, diseases (ICD code) and geographical locations.

Combining kinase inhibitors for optimally co-targeting cancer and drug escape by exploitation of drug target promiscuities.

Cancers resist targeted therapeutics by drug-escape signaling. Multitarget drugs co-targeting cancer and drug-escape mediators (DEMs) are clinically advantageous. DEM coverage may be expanded by drug combinations. This work evaluated to what extent the kinase DEMs (KDEMs) can be optimally co-targeted by drug combinations based on target promiscuities of individual drugs. We focused on 41 approved and 28 clinical trial small molecule kinase inhibitor drugs with available experimental kinome and clinical pharmacokinetic data. From the kinome inhibitory profiles of these drugs, drug combinations were assembled for optimally co-targeting an established cancer target (EGFR, HER2, ABL1, or MEK1) and 9-16 target-associated KDEMs at comparable potency levels as that against the cancer target. Each set of two-, three-, and four-drug combinations co-target 36-71%, 44-89%, 50-88%, and 27-55% KDEMs of EGFR, HER2, ABL1, and MEK1, respectively, compared with the 36, 33, 38, and 18% KDEMs maximally co-targeted by an existing drug or drug combination approved or clinically tested for the respective cancer. Some co-targeted KDEMs are not covered by any existing drug or drug combination. Our work suggested that novel drug combinations may be constructed for optimally co-targeting cancer and drug escape by the exploitation of drug target promiscuities.

NPASS: natural product activity and species source database for natural product research, discovery and tool development.

There has been renewed interests in the exploration of natural products (NPs) for drug discovery, and continuous investigations of the therapeutic claims and mechanisms of traditional and herbal medicines. In-silico methods have been employed for facilitating these studies. These studies and the optimization of in-silico algorithms for NP applications can be facilitated by the quantitative activity and species source data of the NPs. A number of databases collectively provide the structural and other information of ∼470 000 NPs, including qualitative activity information for many NPs, but only ∼4000 NPs are with the experimental activity values. There is a need for the activity and species source data of more NPs. We therefore developed a new database, NPASS (Natural Product Activity and Species Source) to complement other databases by providing the experimental activity values and species sources of 35 032 NPs from 25 041 species targeting 5863 targets (2946 proteins, 1352 microbial species and 1227 cell-lines). NPASS contains 446 552 quantitative activity records (e.g. IC50, Ki, EC50, GI50 or MIC mainly in units of nM) of 222 092 NP-target pairs and 288 002 NP-species pairs. NPASS, http://bidd2.nus.edu.sg/NPASS/, is freely accessible with its contents searchable by keywords, physicochemical property range, structural similarity, species and target search facilities.

Drug sales confirm clinical advantage of multi-target inhibition of drug escapes by anticancer kinase inhibitors.

The clinical advantage of co-targeting cancer drug escape has been indicated by the percentage of these co-targeting drugs among all multi-target drugs in clinics and clinical trials. This clinical advantage needs to be further interrogated from such perspectives as the clinical impact of multi-target inhibition of drug-escape mediators. This impact may be reflected by drug sales data, that is, multi-target inhibition of higher number of drug-escape mediators favors the expanded coverage of drug-resistant patients leading to higher sales. We investigated whether this expectation is followed by the 25 FDA-approved anticancer kinase inhibitors, which were divided into 11 groups of comparable therapeutic mechanisms and approval years. We found 19 (76%) drugs to follow and 3 (12%) drugs not to follow this expectation. The remaining two (8%) and one (4%) drugs cannot be assessed due to insufficient data and incomparability. Therefore, drug sales strongly indicate the clinical advantage of multi-target inhibition of cancer drug escapes.

HEROD: a human ethnic and regional specific omics database.

MOTIVATION: Genetic and gene expression variations within and between populations and across geographical regions have substantial effects on the biological phenotypes, diseases, and therapeutic response. The development of precision medicines can be facilitated by the OMICS studies of the patients of specific ethnicity and geographic region. However, there is an inadequate facility for broadly and conveniently accessing the ethnic and regional specific OMICS data. RESULTS: Here, we introduced a new free database, HEROD, a human ethnic and regional specific OMICS database. Its first version contains the gene expression data of 53 070 patients of 169 diseases in seven ethnic populations from 193 cities/regions in 49 nations curated from the Gene Expression Omnibus (GEO), the ArrayExpress Archive of Functional Genomics Data (ArrayExpress), the Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC). Geographic region information of curated patients was mainly manually extracted from referenced publications of each original study. These data can be accessed and downloaded via keyword search, World map search, and menu-bar search of disease name, the international classification of disease code, geographical region, location of sample collection, ethnic population, gender, age, sample source organ, patient type (patient or healthy), sample type (disease or normal tissue) and assay type on the web interface. AVAILABILITY AND IMPLEMENTATION: The HEROD database is freely accessible at http://bidd2.nus.edu.sg/herod/index.php. The database and web interface are implemented in MySQL, PHP and HTML with all major browsers supported. CONTACT: phacyz@nus.edu.sg.

Kinesin-14 is Important for Chromosome Segregation During Mitosis and Meiosis in the Ciliate Tetrahymena thermophila.

Ciliates such as Tetrahymena thermophila have two distinct nuclei within one cell: the micronucleus that undergoes mitosis and meiosis and the macronucleus that undergoes amitosis, a type of nuclear division that does not involve a bipolar spindle, but still relies on intranuclear microtubules. Ciliates provide an opportunity for the discovery of factors that specifically contribute to chromosome segregation based on a bipolar spindle, by identification of factors that affect the micronuclear but not the macronuclear division. Kinesin-14 is a conserved minus-end directed microtubule motor that cross-links microtubules and contributes to the bipolar spindle sizing and organization. Here, we use homologous DNA recombination to knock out genes that encode kinesin-14 orthologues (KIN141, KIN142) in Tetrahymena. A loss of KIN141 led to severe defects in the chromosome segregation during both mitosis and meiosis but did not affect amitosis. A loss of KIN141 altered the shape of the meiotic spindle in a way consistent with the KIN141's contribution to the organization of the spindle poles. EGFP-tagged KIN141 preferentially accumulated at the spindle poles during the meiotic prophase and metaphase I. Thus, in ciliates, kinesin-14 is important for nuclear divisions that involve a bipolar spindle.

Protrusion-localized STAT3 mRNA promotes metastasis of highly metastatic hepatocellular carcinoma cells in vitro.

AIM: Recent evidence shows that localization of mRNAs and their protein products at cellular protrusions plays a decisive function in the metastasis of cancer cells. The aim of this study was to identify the variety of proteins encoded by protrusion-localized mRNAs and their roles in the metastasis and invasion of liver cancer cells. METHODS: Highly metastatic hepatocellular carcinoma cell line HCCLM3 and non-metastatic hepatocellular carcinoma cell line SMMC-7721 were examined. Cell protrusions (Ps) were separated from cell bodies (CB) using a Boyden chamber assay; total mRNA population in CB and Ps fractions was analyzed using high-throughput direct RNA sequencing. The localization of STAT3 mRNA and protein at Ps was confirmed using RT-qPCR, RNA FISH, and immunofluorescence assays. Cell migration capacity and invasiveness of HCCLM3 cells were evaluated using MTT, wound healing migration and in vitro invasion assays. The interaction between Stat3 and growth factor receptors was explored with co-immunoprecipitation assays. RESULTS: In HCCLM3 cells, 793 mRNAs were identified as being localized in the Ps fraction according to a cut-off value (Ps/CB ratio) >1.6. The Ps-localized mRNAs could be divided into 4 functional groups, and were all closely related to the invasive and metastatic properties. STAT3 mRNA accumulated in the Ps of HCCLM3 cells compared with non-metastatic SMMC-7721 cells. Treatment of HCCLM3 cells with siRNAs against STAT3 mRNA drastically decreased the cell migration and invasion. Moreover, Ps-localized Stat3 was found to interact with pseudopod-enriched platelet-derived growth factor receptor tyrosine kinase (PDGFRTK) in a growth factor-dependent manner. CONCLUSION: This study reveals STAT3 mRNA localization at the Ps of metastatic hepatocellular carcinoma HCCLM3 cells by combining application of genome-wide and gene specific description and functional analysis.

Co-targeting cancer drug escape pathways confers clinical advantage for multi-target anticancer drugs.

Recent investigations have suggested that anticancer therapeutics may be enhanced by co-targeting the primary anticancer target and the corresponding drug escape pathways. Whether this strategy confers statistically significant clinical advantage has not been systematically investigated. This question was probed by the evaluation of the clinical status and the multiple targets of 23 approved and 136 clinical trial multi-target anticancer drugs with particular focus on those co-targeting EGFR, HER2, Abl, VEGFR2, mTOR, PI3K, Alk, MEK, KIT, and DNA topoisomerase, and some of the 14, 7, 13, 20, 6, 5, 7, 2, 4 and 10 cancer drug escape pathways respectively. Most of the approved (73.9%) and phase III (75.0%), the majority of the Phase II (62.8%) and I (53.6%), and the minority of the discontinued (35.3%) multi-target drugs were found to co-target cancer drug escape pathways. This suggests that co-targeting anticancer targets and drug escape pathways confer significant clinical advantage and such strategy can be more extensively explored.

New benzimidazole acridine derivative induces human colon cancer cell apoptosis in vitro via the ROS-JNK signaling pathway.

AIM: To investigate the mechanisms underlying anticancer action of the benzimidazole acridine derivative N-{(1H-benzo[d]imidazol-2-yl)methyl}-2-butylacridin-9-amine(8m) against human colon cancer cells in vitro. METHODS: Human colon cancer cell lines SW480 and HCT116 were incubated in the presence of 8m, and then the cell proliferation and apoptosis were measured. The expression of apoptotic/signaling genes and proteins was detected using RT-PCR and Western blotting. ROS generation and mitochondrial membrane depolarization were visualized with fluorescence microscopy. RESULTS: 8m dose-dependently suppressed the proliferation of SW480 and HCT116 cells with IC50 values of 6.77 and 3.33 µmol/L, respectively. 8m induced apoptosis of HCT116 cells, accompanied by down-regulation of Bcl-2, up-regulation of death receptor-5 (DR5), truncation of Bid, cleavage of PARP, and activation of caspases (including caspase-8 and caspase-9 as well as the downstream caspases-3 and caspase-7). Moreover, 8m selectively activated JNK and p38 without affecting ERK in HCT116 cells. Knockout of JNK1, but not p38, attenuated 8m-induced apoptosis. In addition, 8m induced ROS production and mitochondrial membrane depolarization in HCT116 cells. Pretreatment with the antioxidants N-acetyl cysteine or glutathione attenuated 8m-induced apoptosis and JNK activation in HCT116 cells. CONCLUSION: The new benzimidazole acridine derivative, 8m exerts anticancer activity against human colon cancer cells in vitro by inducing both intrinsic and extrinsic apoptosis pathways via the ROS-JNK1 pathway.

Raman spectroscopic deep learning with signal aggregated representations for enhanced cell phenotype and signature identification.

Feature representation is critical for data learning, particularly in learning spectroscopic data. Machine learning (ML) and deep learning (DL) models learn Raman spectra for rapid, nondestructive, and label-free cell phenotype identification, which facilitate diagnostic, therapeutic, forensic, and microbiological applications. But these are challenged by high-dimensional, unordered, and low-sample spectroscopic data. Here, we introduced novel 2D image-like dual signal and component aggregated representations by restructuring Raman spectra and principal components, which enables spectroscopic DL for enhanced cell phenotype and signature identification. New ConvNet models DSCARNets significantly outperformed the state-of-the-art (SOTA) ML and DL models on six benchmark datasets, mostly with >2% improvement over the SOTA performance of 85-97% accuracies. DSCARNets also performed well on four additional datasets against SOTA models of extremely high performances (>98%) and two datasets without a published supervised phenotype classification model. Explainable DSCARNets identified Raman signatures consistent with experimental indications.

microRNA 21-mediated suppression of Sprouty1 by Pokemon affects liver cancer cell growth and proliferation.

Transcriptional repressor Pokemon is a critical factor in embryogenesis, development, cell proliferation, differentiation, and oncogenesis, thus behaving as an oncogene. Oncomine database suggests a potential correlation between the expressions of Pokemon and Sprouty1. This study investigated the regulatory role of Pokemon in Sprouty1 expression and the effect on liver cancer cell growth and proliferation, revealing a novel miR-21-mediated regulatory circuit. In normal (HL-7702) and cancer (QGY-7703) liver cell lines, Sprouty1 expression is inversely correlated with Pokemon levels. Targeted expression or siRNA-mediated silencing showed that Pokemon is a repressor of Sprouty1 expression at both mRNA and protein levels, but Pokemon cannot affect the promoter activity of Sprouty1. Sprouty1 is a target of miR-21 and interestingly, we found that miR-21 is up-regulated by Pokemon in liver cancer cells. Luciferase reporter assays showed that Pokemon up-regulated miR-21 transcription in a dose-dependent manner, and ChIP assay exhibited a direct binding of Pokemon to the miR-21 promoter at -747 to -399 bp. Site-directed mutagenesis of the GC boxes at -684 to -679 bp and -652 to -647 bp of miR-21 promoter abolished the regulatory activity by Pokemon. Furthermore, we found that the modulation of Pokemon and miR-21 expression affected the growth and proliferation of liver cancer cells QGY-7703. In summary, our findings demonstrate that Pokemon suppresses Sprouty1 expression through a miR-21-mediated mechanism, affecting the growth and proliferation of liver cancer cells. This study recognized miR-21 and Sprouty1 as novel targets of the Pokemon regulatory network.

Homeostatic regulatory role of Pokemon in NF-κB signaling: stimulating both p65 and IκBα expression in human hepatocellular carcinoma cells.

NF-κB consists of p50, p65 (RelA), p52, c-Rel, and RelB, and among them p65 is a representative protein to investigate the regulation and function of this signaling. NF-κB integrates inflammation and carcinogenesis and regulates the expression of a variety of genes in response to immunity, inflammation, and apoptosis. IκBα acts as an inhibitor of NF-κB through forming an inactive NF-κB/IκBα complex. Pokemon is a ubiquitous transcription factor involved in different signaling pathways, playing a pivotal role in cell proliferation, anti-apoptosis, embryonic development, and maintenance. In this study, we found that p65 and IκBα are both novel regulatory targets of Pokemon. Ectopic expression of Pokemon in immortalized liver cells HL7702 enhanced p65 and IκBα expression, whereas silencing of Pokemon in hepatocellular carcinoma cells QGY7703 reduced cellular p65 levels. ChIP assay and targeted mutagenesis revealed that Pokemon directly binds to the element of -434 to -430 bp in p65 promoter and of -453 to -448 bp in IκBα promoter and stimulates luciferase reporter gene expression. Co-transfection of Pokemon with p65 or IκBα promoter-reporter notably enhanced their promoter activity. These data suggest that Pokemon activates the expression of both p65 and IκBα by sequence-specific binding to their promoters and plays a dual role in regulating NF-κB signaling.

Escherichia fergusonii bacteremia after esophagectomy: a case report and literature review.

We report the case of a 68-year-old man who experienced Escherichia fergusonii bacteremia after esophageal cancer surgery. The patient presented with complaints of abdominal pain persisting for 1 week. The patient was diagnosed with esophageal malignancy, which was confirmed by surgical exploration and pathological biopsy. The patient developed septic shock on postoperative day 12, and blood culture suggested the growth of E. fergusonii. After treatment with meropenem, the patient's clinical symptoms improved significantly, and the second culture was negative. In this paper, we discuss the characteristics, diagnosis, and treatment of E. fergusonii. E. fergusonii is rarely reported, and its pathogenesis, drug resistance, and potential effects have not been completely confirmed. Thus, this case report adds valuable knowledge to the literature on E. fergusonii.

Enhanced metagenomic deep learning for disease prediction and consistent signature recognition by restructured microbiome 2D representations.

Metagenomic analysis has been explored for disease diagnosis and biomarker discovery. Low sample sizes, high dimensionality, and sparsity of metagenomic data challenge metagenomic investigations. Here, an unsupervised microbial embedding, grouping, and mapping algorithm (MEGMA) was developed to transform metagenomic data into individualized multichannel microbiome 2D representation by manifold learning and clustering of microbial profiles (e.g., composition, abundance, hierarchy, and taxonomy). These 2D representations enable enhanced disease prediction by established ConvNet-based AggMapNet models, outperforming the commonly used machine learning and deep learning models in metagenomic benchmark datasets. These 2D representations combined with AggMapNet explainable module robustly identified more reliable and replicable disease-prediction microbes (biomarkers). Employing the MEGMA-AggMapNet pipeline for biomarker identification from 5 disease datasets, 84% of the identified biomarkers have been described in over 74 distinct works as important for these diseases. Moreover, the method also discovered highly consistent sets of biomarkers in cross-cohort colorectal cancer (CRC) patients and microbial shifts in different CRC stages.

Deep learning of 2D-Restructured gene expression representations for improved low-sample therapeutic response prediction.

Clinical outcome prediction is important for stratified therapeutics. Machine learning (ML) and deep learning (DL) methods facilitate therapeutic response prediction from transcriptomic profiles of cells and clinical samples. Clinical transcriptomic DL is challenged by the low-sample sizes (34-286 subjects), high-dimensionality (up to 21,653 genes) and unordered nature of clinical transcriptomic data. The established methods rely on ML algorithms at accuracy levels of 0.6-0.8 AUC/ACC values. Low-sample DL algorithms are needed for enhanced prediction capability. Here, an unsupervised manifold-guided algorithm was employed for restructuring transcriptomic data into ordered image-like 2D-representations, followed by efficient DL of these 2D-representations with deep ConvNets. Our DL models significantly outperformed the state-of-the-art (SOTA) ML models on 82% of 17 low-sample benchmark datasets (53% with >0.05 AUC/ACC improvement). They are more robust than the SOTA models in cross-cohort prediction tasks, and in identifying robust biomarkers and response-dependent variational patterns consistent with experimental indications.