Large-Scale Machine Learning Analysis Reveals DNA Methylation and Gene Expression Response Signatures for Gemcitabine-Treated Pancreatic Cancer.

Background: Gemcitabine is a first-line chemotherapy for pancreatic adenocarcinoma (PAAD), but many PAAD patients do not respond to gemcitabine-containing treatments. Being able to predict such nonresponders would hence permit the undelayed administration of more promising treatments while sparing gemcitabine life-threatening side effects for those patients. Unfortunately, the few predictors of PAAD patient response to this drug are weak, none of them exploiting yet the power of machine learning (ML). Methods: Here, we applied ML to predict the response of PAAD patients to gemcitabine from the molecular profiles of their tumors. More concretely, we collected diverse molecular profiles of PAAD patient tumors along with the corresponding clinical data (gemcitabine responses and clinical features) from the Genomic Data Commons resource. From systematically combining 8 tumor profiles with 16 classification algorithms, each of the resulting 128 ML models was evaluated by multiple 10-fold cross-validations. Results: Only 7 of these 128 models were predictive, which underlines the importance of carrying out such a large-scale analysis to avoid missing the most predictive models. These were here random forest using 4 selected mRNAs [0.44 Matthews correlation coefficient (MCC), 0.785 receiver operating characteristic-area under the curve (ROC-AUC)] and XGBoost combining 12 DNA methylation probes (0.32 MCC, 0.697 ROC-AUC). By contrast, the hENT1 marker obtained much worse random-level performance (practically 0 MCC, 0.5 ROC-AUC). Despite not being trained to predict prognosis (overall and progression-free survival), these ML models were also able to anticipate this patient outcome. Conclusions: We release these promising ML models so that they can be evaluated prospectively on other gemcitabine-treated PAAD patients.

NKG2D Fine-Tunes the Local Inflammatory Response in Colorectal Cancer.

Treating colorectal cancer (CRC) is a major challenge due to the heterogeneous immunological, clinical and pathological landscapes. Immunotherapy has so far only proven effective in a very limited subgroup of CRC patients. To better define the immune landscape, we examined the immune gene expression profile in various subsets of CRC patients and used a mouse model of intestinal tumors to dissect immune functions. We found that the NK cell receptor, natural-killer group 2 member D (NKG2D, encoded by KLRK1) and NKG2D ligand gene expression is elevated in the most immunogenic subset of CRC patients. High level of KLRK1 positively correlated with the mRNA expression of IFNG and associated with a poor survival of CRC patients. We further show that NKG2D deficiency in the Apcmin/+ mouse model of intestinal tumorigenesis led to reduced intratumoral IFNγ production, reduced tumorigenesis and enhanced survival, suggesting that the high levels of IFNγ observed in the tumors of CRC patients may be a consequence of NKG2D engagement. The mechanisms governing the contribution of NKG2D to CRC progression highlighted in this study will fuel discussions about (i) the benefit of targeting NKG2D in CRC patients and (ii) the need to define the predictive value of NKG2D and NKG2D ligand expression across tumor types.

Interpretable Machine Learning Models to Predict the Resistance of Breast Cancer Patients to Doxorubicin from Their microRNA Profiles.

Doxorubicin is a common treatment for breast cancer. However, not all patients respond to this drug, which sometimes causes life-threatening side effects. Accurately anticipating doxorubicin-resistant patients would therefore permit to spare them this risk while considering alternative treatments without delay. Stratifying patients based on molecular markers in their pretreatment tumors is a promising approach to advance toward this ambitious goal, but single-gene gene markers such as HER2 expression have not shown to be sufficiently predictive. The recent availability of matched doxorubicin-response and diverse molecular profiles across breast cancer patients permits now analysis at a much larger scale. 16 machine learning algorithms and 8 molecular profiles are systematically evaluated on the same cohort of patients. Only 2 of the 128 resulting models are substantially predictive, showing that they can be easily missed by a standard-scale analysis. The best model is classification and regression tree (CART) nonlinearly combining 4 selected miRNA isoforms to predict doxorubicin response (median Matthew correlation coefficient (MCC) and area under the curve (AUC) of 0.56 and 0.80, respectively). By contrast, HER2 expression is significantly less predictive (median MCC and AUC of 0.14 and 0.57, respectively). As the predictive accuracy of this CART model increases with larger training sets, its update with future data should result in even better accuracy.

Predicting Cancer Drug Response In Vivo by Learning an Optimal Feature Selection of Tumour Molecular Profiles.

(1) Background: Inter-tumour heterogeneity is one of cancer's most fundamental features. Patient stratification based on drug response prediction is hence needed for effective anti-cancer therapy. However, single-gene markers of response are rare and/or may fail to achieve a significant impact in the clinic. Machine Learning (ML) is emerging as a particularly promising complementary approach to precision oncology. (2) Methods: Here we leverage comprehensive Patient-Derived Xenograft (PDX) pharmacogenomic data sets with dimensionality-reducing ML algorithms with this purpose. (3) Results: Combining multiple gene alterations via ML leads to better discrimination between sensitive and resistant PDXs in 19 of the 26 analysed cases. Highly predictive ML models employing concise gene lists were found for three cases: paclitaxel (breast cancer), binimetinib (breast cancer) and cetuximab (colorectal cancer). Interestingly, each of these multi-gene ML models identifies some treatment-responsive PDXs not harbouring the best actionable mutation for that case. Thus, ML multi-gene predictors generally have much fewer false negatives than the corresponding single-gene marker. (4) Conclusions: As PDXs often recapitulate clinical outcomes, these results suggest that many more patients could benefit from precision oncology if ML algorithms were also applied to existing clinical pharmacogenomics data, especially those algorithms generating classifiers combining data-selected gene alterations.

A gentle introduction to understanding preclinical data for cancer pharmaco-omic modeling.

A central goal of precision oncology is to administer an optimal drug treatment to each cancer patient. A common preclinical approach to tackle this problem has been to characterize the tumors of patients at the molecular and drug response levels, and employ the resulting datasets for predictive in silico modeling (mostly using machine learning). Understanding how and why the different variants of these datasets are generated is an important component of this process. This review focuses on providing such introduction aimed at scientists with little previous exposure to this research area.

NF-κB-dependent IRF1 activation programs cDC1 dendritic cells to drive antitumor immunity.

Conventional type 1 dendritic cells (cDC1s) are critical for antitumor immunity. They acquire antigens from dying tumor cells and cross-present them to CD8+ T cells, promoting the expansion of tumor-specific cytotoxic T cells. However, the signaling pathways that govern the antitumor functions of cDC1s in immunogenic tumors are poorly understood. Using single-cell transcriptomics to examine the molecular pathways regulating intratumoral cDC1 maturation, we found nuclear factor κB (NF-κB) and interferon (IFN) pathways to be highly enriched in a subset of functionally mature cDC1s. We identified an NF-κB-dependent and IFN-γ-regulated gene network in cDC1s, including cytokines and chemokines specialized in the recruitment and activation of cytotoxic T cells. By mapping the trajectory of intratumoral cDC1 maturation, we demonstrated the dynamic reprogramming of tumor-infiltrating cDC1s by NF-κB and IFN signaling pathways. This maturation process was perturbed by specific inactivation of either NF-κB or IFN regulatory factor 1 (IRF1) in cDC1s, resulting in impaired expression of IFN-γ-responsive genes and consequently a failure to efficiently recruit and activate antitumoral CD8+ T cells. Last, we demonstrate the relevance of these findings to patients with melanoma, showing that activation of the NF-κB/IRF1 axis in association with cDC1s is linked with improved clinical outcome. The NF-κB/IRF1 axis in cDC1s may therefore represent an important focal point for the development of new diagnostic and therapeutic approaches to improve cancer immunotherapy.

Recent progress on the prospective application of machine learning to structure-based virtual screening.

As more bioactivity and protein structure data become available, scoring functions (SFs) using machine learning (ML) to leverage these data sets continue to gain further accuracy and broader applicability. Advances in our understanding of the optimal ways to train and evaluate these ML-based SFs have introduced further improvements. One of these advances is how to select the most suitable decoys (molecules assumed inactive) to train or test an ML-based SF on a given target. We also review the latest applications of ML-based SFs for prospective structure-based virtual screening (SBVS), with a focus on the observed improvement over those using classical SFs. Finally, we provide recommendations for future prospective SBVS studies based on the findings of recent methodological studies.

Characterizing the Relationship Between the Chemical Structures of Drugs and their Activities on Primary Cultures of Pediatric Solid Tumors.

BACKGROUND: Despite continued efforts to develop new treatments, there is an urgent need to discover new drug leads to treat tumors exhibiting primary or secondary resistance to existing drugs. Cell cultures derived from patient-derived orthotopic xenografts are promising pre-clinical models to better predict drug response in cancer recurrence. OBJECTIVE: The aim of the study was to investigate the relationship between the physiochemical properties of drugs and their in vitro potency as well as identifying chemical scaffolds biasedtowards selectivity or promiscuity of such drugs. METHODS: The bioactivities of 158 drugs screened against cell cultures derived from 30 cancer orthotopic patient-derived xenograft (O-PDX) models were considered. Drugs were represented by physicochemical descriptors and chemical structure fingerprints. Supervised learning was employed to model the relationship between features and in vitro potency. RESULTS: Drugs with in vitro potency for alveolar rhabdomyosarcoma and osteosarcoma tend to have a higher number of rings, two carbon-hetero bonds and halogens. Selective and promiscuous scaffolds for these phenotypic targets were identified. Highly-predictive models of in vitro potency were obtained across these 30 targets, which can be applied to unseen molecules via a webserver (https://rnewbie.shinyapps.io/Shobek-master). CONCLUSION: It is possible to identify privileged chemical scaffolds and predict the in vitro potency of unseen molecules across these 30 targets This information and models should be helpful to select which molecules to screen against these primary cultures of pediatric solid tumors.

Identification and Validation of Carbonic Anhydrase II as the First Target of the Anti-Inflammatory Drug Actarit.

Background and purpose: Identifying the macromolecular targets of drug molecules is a fundamental aspect of drug discovery and pharmacology. Several drugs remain without known targets (orphan) despite large-scale in silico and in vitro target prediction efforts. Ligand-centric chemical-similarity-based methods for in silico target prediction have been found to be particularly powerful, but the question remains of whether they are able to discover targets for target-orphan drugs. Experimental Approach: We used one of these in silico methods to carry out a target prediction analysis for two orphan drugs: actarit and malotilate. The top target predicted for each drug was carbonic anhydrase II (CAII). Each drug was therefore quantitatively evaluated for CAII inhibition to validate these two prospective predictions. Key Results: Actarit showed in vitro concentration-dependent inhibition of CAII activity with submicromolar potency (IC50 = 422 nM) whilst no consistent inhibition was observed for malotilate. Among the other 25 targets predicted for actarit, RORγ (RAR-related orphan receptor-gamma) is promising in that it is strongly related to actarit's indication, rheumatoid arthritis (RA). Conclusion and Implications: This study is a proof-of-concept of the utility of MolTarPred for the fast and cost-effective identification of targets of orphan drugs. Furthermore, the mechanism of action of actarit as an anti-RA agent can now be re-examined from a CAII-inhibitor perspective, given existing relationships between this target and RA. Moreover, the confirmed CAII-actarit association supports investigating the repositioning of actarit on other CAII-linked indications (e.g., hypertension, epilepsy, migraine, anemia and bone, eye and cardiac disorders).

MolTarPred: A web tool for comprehensive target prediction with reliability estimation.

Molecular target prediction can provide a starting point to understand the efficacy and side effects of phenotypic screening hits. Unfortunately, the vast majority of in silico target prediction methods are not available as web tools. Furthermore, these are limited in the number of targets that can be predicted, do not estimate which target predictions are more reliable and/or lack comprehensive retrospective validations. We present MolTarPred ( http://moltarpred.marseille.inserm.fr/), a user-friendly web tool for predicting protein targets of small organic compounds. It is powered by a large knowledge base comprising 607,659 compounds and 4,553 macromolecular targets collected from the ChEMBL database. In about 1 min, the predicted targets for the supplied molecule will be listed in a table. The chemical structures of the query molecule and the most similar compounds annotated with the predicted target will also be shown to permit visual inspection and comparison. Practical examples of the use of MolTarPred are showcased. MolTarPred is a new resource for scientists that require a more complete knowledge of the polypharmacology of a molecule. The introduction of a reliability score constitutes an attractive functionality of MolTarPred, as it permits focusing experimental confirmatory tests on the most reliable predictions, which leads to higher prospective hit rates.

Autophagy in dendritic cells.

Autophagy and immunity share the property of being auto-protective for the organism. Autophagy is an important degradation pathway that buffers nutrient deprivation by recycling macromolecules in organisms from yeast to man. Perturbations in autophagy are associated with inflammation and cancer development. Emerging studies have characterized the molecular details regarding how autophagy is controlled by immune cells. Among these, dendritic cells (DCs) are one of the most potent professional antigen-presenting cells critical for the activation of naïve T cells to maintain immune tolerance and drive protective immunity to infection and cancer. DCs undergo functional maturation that can either lead to an immunostimulatory phenotype, as in the context of infection, or to a tolerogenic phenotype associated with immunosuppression to self-antigens, as well as to cancer. An increasing number of recent studies has characterized the involvement of autophagy in DC functions in various physiological and pathological contexts. Here, we provide a comprehensive review of these outcomes and discuss the limitation of the models used and the forefront of the knowledge concerning the crosstalk between autophagy and DC biology.

Transcriptional regulation of mammalian autophagy at a glance.

Macroautophagy, hereafter referred to as autophagy, is a catabolic process that results in the lysosomal degradation of cytoplasmic contents ranging from abnormal proteins to damaged cell organelles. It is activated  under diverse conditions, including nutrient deprivation and hypoxia. During autophagy, members of the core autophagy-related (ATG) family of proteins mediate membrane rearrangements, which lead to the engulfment and degradation of cytoplasmic cargo. Recently, the nuclear regulation of autophagy, especially by transcription factors and histone modifiers, has gained increased attention. These factors are not only involved in rapid responses to autophagic stimuli, but also regulate the long-term outcome of autophagy. Now there are more than 20 transcription factors that have been shown to be linked to the autophagic process. However, their interplay and timing appear enigmatic as several have been individually shown to act as major regulators of autophagy. This Cell Science at a Glance article and the accompanying poster highlights the main cellular regulators of transcription involved in mammalian autophagy and their target genes.

Mammalian Autophagy: How Does It Work?

Autophagy is a conserved intracellular pathway that delivers cytoplasmic contents to lysosomes for degradation via double-membrane autophagosomes. Autophagy substrates include organelles such as mitochondria, aggregate-prone proteins that cause neurodegeneration and various pathogens. Thus, this pathway appears to be relevant to the pathogenesis of diverse diseases, and its modulation may have therapeutic value. Here, we focus on the cell and molecular biology of mammalian autophagy and review the key proteins that regulate the process by discussing their roles and how these may be modulated by posttranslational modifications. We consider the membrane-trafficking events that impact autophagy and the questions relating to the sources of autophagosome membrane(s). Finally, we discuss data from structural studies and some of the insights these have provided.

Transcriptional regulation of Annexin A2 promotes starvation-induced autophagy.

Autophagy is an important degradation pathway, which is induced after starvation, where it buffers nutrient deprivation by recycling macromolecules in organisms from yeast to man. While the classical pathway mediating this response is via mTOR inhibition, there are likely to be additional pathways that support the process. Here, we identify Annexin A2 as an autophagy modulator that regulates autophagosome formation by enabling appropriate ATG9A trafficking from endosomes to autophagosomes via actin. This process is dependent on the Annexin A2 effectors ARP2 and Spire1. Annexin A2 expression increases after starvation in cells in an mTOR-independent fashion. This is mediated via Jun N-terminal kinase activation of c-Jun, which, in turn, enhances the trans-activation of the Annexin A2 promoter. Annexin A2 knockdown abrogates starvation-induced autophagy, while its overexpression induces autophagy. Hence, c-Jun-mediated transcriptional responses support starvation-induced autophagy by regulating Annexin A2 expression levels.

XIAP and cIAP1 amplifications induce Beclin 1-dependent autophagy through NFκB activation.

Perturbations in autophagy and apoptosis are associated with cancer development. XIAP and cIAP1 are two members of the inhibitors of apoptosis protein family whose expression is elevated in different cancers. Here we report that XIAP and cIAP1 induce autophagy by upregulating the transcription of Beclin 1, an essential autophagy gene. The E3 ubiquitin ligase activity of both proteins activates NFκB signalling, leading to the direct binding of p65 to the promoter of Beclin 1 and to its transcriptional activation. This mechanism may be relevant in cancer cells, since we found increased levels of autophagy in different B-cell lymphoma-derived cell lines where XIAP is overexpressed and pharmacological inhibition of XIAP in these cell lines reduced autophagosome biogenesis. Thus, the chemotherapy resistance associated with XIAP and cIAP1 overexpression observed in several human cancers may be, at least in part, due to the Beclin 1-dependent autophagy activation by IAPs described in this study. In this context, the disruption of this increased autophagy might represent a valuable pharmacological tool to be included in combined anti-neoplastic therapies.

PTEN increases autophagy and inhibits the ubiquitin-proteasome pathway in glioma cells independently of its lipid phosphatase activity.

Two major mechanisms of intracellular protein degradation, autophagy and the ubiquitin-proteasome pathway, operate in mammalian cells. PTEN, which is frequently mutated in glioblastomas, is a tumor suppressor gene that encodes a dual specificity phosphatase that antagonizes the phosphatidylinositol 3-kinase class I/AKT/mTOR pathway, which is a key regulator of autophagy. Here, we investigated in U87MG human glioma cells the role of PTEN in the regulation of autophagy and the ubiquitin-proteasome pathway, because both are functionally linked and are relevant in cancer progression. Since U87MG glioma cells lack a functional PTEN, we used stable clones that express, under the control of a tetracycline-inducible system (Tet-on), wild-type PTEN and two of its mutants, G129E-PTEN and C124S-PTEN, which, respectively, lack the lipid phosphatase activity only and both the lipid and the protein phosphatase activities of this protein. Expression of PTEN in U87MG glioma cells decreased proteasome activity and also reduced protein ubiquitination. On the contrary, expression of PTEN increased the autophagic flux and the lysosomal mass. Interestingly, and although PTEN negatively regulates the phosphatidylinositol 3-kinase class I/AKT/mTOR signaling pathway by its lipid phosphatase activity, both effects in U87MG cells were independent of this activity. These results suggest a new mTOR-independent signaling pathway by which PTEN can regulate in opposite directions the main mechanisms of intracellular protein degradation.

Ca²âº-sensor proteins in the autophagic and endocytic traffic.

Autophagy and endocytosis are two evolutionarily conserved catabolic processes that comprise vesicle trafficking events for the clearance of the sequestered intracellular and extracellular cargo. Both start differently but end in the same compartment, the lysosome. Mounting evidences from the last years have established the involvement of proteins sensitive to intracellular Ca(2+) in the control of the early autophagic steps and in the traffic of autophagic, endocytic and lysosomal vesicles. However, this knowledge is based on dispersed outcomes that do not set up a consensus model of the Ca(2+)-dependent control of autophagy and endocytosis. Here, we will provide a critical synopsis of insights from the last decade on the involvement of Ca(2+)-sensor proteins in the activation of autophagy and in fusion events of endocytic vesicles, autophagosomes and lysosomes.

New Ca(2+)-dependent regulators of autophagosome maturation.

Autophagy is a membrane trafficking pathway responsible for the breakdown of unwanted intracellular materials and crucial for the cell healthiness and survival. In the autophagic flux, various dynamic membrane rearrangements occurs starting with the elongation of the phagophore and its closure to build an autophagosome and ending with its fusion with late endosomes and lysosomes to form an autolysosome. Although Ca(2+) is a well established regulator of membrane fusion events, little is known about its role in these processes during autophagy. Recent studies, based on proteomic analyses of lysosomal membranes, have provided new insights into this field of study. Thus, the levels on lysosomal membranes of annexin A1, annexin A5 and copine 1, three proteins that bind to phospholipid membranes in a Ca(2+)-dependent manner, increased under nutrient deprivation, a condition that promotes autophagic degradation. In addition, two different studies showed that annexin A5 and annexin A1 are involved in autophagosome maturation. Here, we discuss the molecular mechanisms by which the fusion of autophagosomes with endosomes and lysosomes could be regulated by these three proteins and Ca(2+).

Withdrawal of essential amino acids increases autophagy by a pathway involving Ca2+/calmodulin-dependent kinase kinase-ß (CaMKK-ß).

Autophagy is the main lysosomal catabolic process that becomes activated under stress conditions, such as amino acid starvation and cytosolic Ca(2+) upload. However, the molecular details on how both conditions control autophagy are still not fully understood. Here we link essential amino acid starvation and Ca(2+) in a signaling pathway to activate autophagy. We show that withdrawal of essential amino acids leads to an increase in cytosolic Ca(2+), arising from both extracellular medium and intracellular stores, which induces the activation of adenosine monophosphate-activated protein kinase (AMPK) via Ca(2+)/calmodulin-dependent kinase kinase-ß (CaMKK-ß). Furthermore, we show that autophagy induced by amino acid starvation requires AMPK, as this induction is attenuated in its absence. Subsequently, AMPK activates UNC-51-like kinase (ULK1), a mammalian autophagy-initiating kinase, through phosphorylation at Ser-555 in a process that requires CaMKK-ß. Finally, the mammalian target of rapamycin complex C1 (mTORC1), a negative regulator of autophagy downstream of AMPK, is inhibited by amino acid starvation in a Ca(2+)-sensitive manner, and CaMKK-ß appears to be important for mTORC1 inactivation, especially in the absence of extracellular Ca(2+). All these results highlight that amino acid starvation regulates autophagy in part through an increase in cellular Ca(2+) that activates a CaMKK-ß-AMPK pathway and inhibits mTORC1, which results in ULK1 stimulation.

Annexin A5 stimulates autophagy and inhibits endocytosis.

Macroautophagy is a major lysosomal catabolic process activated particularly under starvation in eukaryotic cells. A new organelle, the autophagosome, engulfs cytoplasmic substrates, which are degraded after fusion with endosomes and/or lysosomes. During a shotgun proteome analysis of purified lysosomal membranes from mouse fibroblasts, a Ca(2+)-dependent phospholipid-binding protein, annexin A5, was found to increase on lysosomal membranes under starvation. This suggests a role for this protein, an abundant annexin with a still unknown intracellular function, in starvation-induced lysosomal degradation. Transient overexpression and silencing experiments showed that annexin A5 increased lysosomal protein degradation, and colocalisation experiments, based on GFP sensitivity to lysosomal acidic pH, indicated that this was mainly the result of inducing autophagosome-lysosome fusion. Annexin A5 also inhibited the endocytosis of a fluid-phase marker and cholera toxin, but not receptor-mediated endocytosis. Therefore, we propose a double and opposite role of annexin A5 in regulating the endocytic and autophagic pathways and the fusion of autophagosomes with lysosomes and endosomes.