Lymph node and tumor-associated PD-L1+ macrophages antagonize dendritic cell vaccines by suppressing CD8+ T cells

Current immunotherapies provide limited benefits against T cell-depleted tumors, calling for therapeutic innovation. Using multi-omics integration of cancer patient data, we predict a type I interferon (IFN) responseHIGH state of dendritic cell (DC) vaccines, with efficacious clinical impact. However, preclinical DC vaccines recapitulating this state by combining immunogenic cancer cell death with induction of type I IFN responses fail to regress mouse tumors lacking T cell infiltrates. Here, in lymph nodes (LNs), instead of activating CD4+/CD8+ T cells, DCs stimulate immunosuppressive programmed death-ligand 1-positive (PD-L1+) LN-associated macrophages (LAMs). Moreover, DC vaccines also stimulate PD-L1+ tumor-associated macrophages (TAMs). This creates two anatomically distinct niches of PD-L1+ macrophages that suppress CD8+ T cells. Accordingly, a combination of PD-L1 blockade with DC vaccines achieves significant tumor regression by depleting PD-L1+ macrophages, suppressing myeloid inflammation, and de-inhibiting effector/stem-like memory T cells. Importantly, clinical DC vaccines also potentiate T cell-suppressive PD-L1+ TAMs in glioblastoma patients. We propose that a multimodal immunotherapy and vaccination regimen is mandatory to overcome T cell-depleted tumors.

The cell stress and immunity cycle in cancer: Toward next generation of cancer immunotherapy

The cellular stress and immunity cycle is a cornerstone of organismal homeostasis. Stress activates intracellular and intercellular communications within a tissue or organ to initiate adaptive responses aiming to resolve the origin of this stress. If such local measures are unable to ameliorate this stress, then intercellular communications expand toward immune activation with the aim of recruiting immune cells to effectively resolve the situation while executing tissue repair to ameliorate any damage and facilitate homeostasis. This cellular stress-immunity cycle is severely dysregulated in diseased contexts like cancer. On one hand, cancer cells dysregulate the normal cellular stress responses to reorient them toward upholding growth at all costs, even at the expense of organismal integrity and homeostasis. On the other hand, the tumors severely dysregulate or inhibit various components of organismal immunity, for example, by facilitating immunosuppressive tumor landscape, lowering antigenicity, and increasing T-cell dysfunction. In this review we aim to comprehensively discuss the basis behind tumoral dysregulation of cellular stress-immunity cycle. We also offer insights into current understanding of the regulators and deregulators of this cycle and how they can be targeted for conceptualizing successful cancer immunotherapy regimen.

Single cell dynamics of tumor specificity vs bystander activity in CD8+ T cells define the diverse immune landscapes in colorectal cancer

CD8+ T cell activation via immune checkpoint blockade (ICB) is successful in microsatellite instable (MSI) colorectal cancer (CRC) patients. By comparison, the success of immunotherapy against microsatellite stable (MSS) CRC is limited. Little is known about the most critical features of CRC CD8+ T cells that together determine the diverse immune landscapes and contrasting ICB responses. Hence, we pursued a deep single cell mapping of CRC CD8+ T cells on transcriptomic and T cell receptor (TCR) repertoire levels in a diverse patient cohort, with additional surface proteome validation. This revealed that CRC CD8+ T cell dynamics are underscored by complex interactions between interferon-γ signaling, tumor reactivity, TCR repertoire, (predicted) TCR antigen-specificities, and environmental cues like gut microbiome or colon tissue-specific 'self-like' features. MSI CRC CD8+ T cells showed tumor-specific activation reminiscent of canonical 'T cell hot' tumors, whereas the MSS CRC CD8+ T cells exhibited tumor unspecific or bystander-like features. This was accompanied by inflammation reminiscent of 'pseudo-T cell hot' tumors. Consequently, MSI and MSS CRC CD8+ T cells showed overlapping phenotypic features that differed dramatically in their TCR antigen-specificities. Given their high discriminating potential for CD8+ T cell features/specificities, we used the single cell tumor-reactive signaling modules in CD8+ T cells to build a bulk tumor transcriptome classification for CRC patients. This "Immune Subtype Classification" (ISC) successfully distinguished various tumoral immune landscapes that showed prognostic value and predicted immunotherapy responses in CRC patients. Thus, we deliver a unique map of CRC CD8+ T cells that drives a novel tumor immune landscape classification, with relevance for immunotherapy decision-making.

Obesity-associated changes in molecular biology of primary breast cancer

Obesity is associated with an increased risk of developing breast cancer (BC) and worse prognosis in BC patients, yet its impact on BC biology remains understudied in humans. This study investigates how the biology of untreated primary BC differs according to patients' body mass index (BMI) using data from >2,000 patients. We identify several genomic alterations that are differentially prevalent in overweight or obese patients compared to lean patients. We report evidence supporting an ageing accelerating effect of obesity at the genetic level. We show that BMI-associated differences in bulk transcriptomic profile are subtle, while single cell profiling allows detection of more pronounced changes in different cell compartments. These analyses further reveal an elevated and unresolved inflammation of the BC tumor microenvironment associated with obesity, with distinct characteristics contingent on the estrogen receptor status. Collectively, our analyses imply that obesity is associated with an inflammaging-like phenotype. We conclude that patient adiposity may play a significant role in the heterogeneity of BC and should be considered for BC treatment tailoring.

Tumour immune rejection triggered by activation of α2-adrenergic receptors

Immunotherapy based on immunecheckpoint blockade (ICB) using antibodies induces rejection of tumours and brings clinical benefit in patients with various cancer types¹. However, tumours often resist immune rejection. Ongoing efforts trying to increase tumour response rates are based on combinations of ICB with compounds that aim to reduce immunosuppression in the tumour microenvironment but usually have little effect when used as monotherapies^2,3. Here we show that agonists of α2-adrenergic receptors (α2-AR) have very strong anti-tumour activity when used as monotherapies in multiple immunocompetent tumour models, including ICB-resistant models, but not in immunodeficient models. We also observed marked effects in human tumour xenografts implanted in mice reconstituted with human lymphocytes. The anti-tumour effects of α2-AR agonists were reverted by α2-AR antagonists, and were absent in Adra2a-knockout (encoding α2a-AR) mice, demonstrating on-target action exerted on host cells, not tumour cells. Tumours from treated mice contained increased infiltrating T lymphocytes and reduced myeloid suppressor cells, which were more apoptotic. Single-cell RNA-sequencing analysis revealed upregulation of innate and adaptive immune response pathways in macrophages and T cells. To exert their anti-tumour effects, α2-AR agonists required CD4⁺ T lymphocytes, CD8⁺ T lymphocytes and macrophages. Reconstitution studies in Adra2a-knockout mice indicated that the agonists acted directly on macrophages, increasing their ability to stimulate T lymphocytes. Our results indicate that α2-AR agonists, some of which are available clinically, could substantially improve the clinical efficacy of cancer immunotherapy.

Trial watch: chemotherapy-induced immunogenic cell death in oncology

Immunogenic cell death (ICD) refers to an immunologically distinct process of regulated cell death that activates, rather than suppresses, innate and adaptive immune responses. Such responses culminate into T cell-driven immunity against antigens derived from dying cancer cells. The potency of ICD is dependent on the immunogenicity of dying cells as defined by the antigenicity of these cells and their ability to expose immunostimulatory molecules like damage-associated molecular patterns (DAMPs) and cytokines like type I interferons (IFNs). Moreover, it is crucial that the host's immune system can adequately detect the antigenicity and adjuvanticity of these dying cells. Over the years, several well-known chemotherapies have been validated as potent ICD inducers, including (but not limited to) anthracyclines, paclitaxels, and oxaliplatin. Such ICD-inducing chemotherapeutic drugs can serve as important combinatorial partners for anti-cancer immunotherapies against highly immuno-resistant tumors. In this Trial Watch, we describe current trends in the preclinical and clinical integration of ICD-inducing chemotherapy in the existing immuno-oncological paradigms.

Tryptophan depletion sensitizes the AHR pathway by increasing AHR expression and GCN2/LAT1-mediated kynurenine uptake, and potentiates induction of regulatory T lymphocytes

BACKGROUND

Indoleamine 2,3-dioxygenase 1 (IDO1) and tryptophan-dioxygenase (TDO) are enzymes catabolizing the essential amino acid tryptophan into kynurenine. Expression of these enzymes is frequently observed in advanced-stage cancers and is associated with poor disease prognosis and immune suppression. Mechanistically, the respective roles of tryptophan shortage and kynurenine production in suppressing immunity remain unclear. Kynurenine was proposed as an endogenous ligand for the aryl hydrocarbon receptor (AHR), which can regulate inflammation and immunity. However, controversy remains regarding the role of AHR in IDO1/TDO-mediated immune suppression, as well as the involvement of kynurenine. In this study, we aimed to clarify the link between IDO1/TDO expression, AHR pathway activation and immune suppression.

METHODS

AHR expression and activation was analyzed by RT-qPCR and western blot analysis in cells engineered to express IDO1/TDO, or cultured in medium mimicking tryptophan catabolism by IDO1/TDO. In vitro differentiation of naïve CD4⁺ T cells into regulatory T cells (Tregs) was compared in T cells isolated from mice bearing different Ahr alleles or a knockout of Ahr, and cultured in medium with or without tryptophan and kynurenine.

RESULTS

We confirmed that IDO1/TDO expression activated AHR in HEK-293-E cells, as measured by the induction of AHR target genes. Unexpectedly, AHR was also overexpressed on IDO1/TDO expression. AHR overexpression did not depend on kynurenine but was triggered by tryptophan deprivation. Multiple human tumor cell lines overexpressed AHR on tryptophan deprivation. AHR overexpression was not dependent on general control non-derepressible 2 (GCN2), and strongly sensitized the AHR pathway. As a result, kynurenine and other tryptophan catabolites, which are weak AHR agonists in normal conditions, strongly induced AHR target genes in tryptophan-depleted conditions. Tryptophan depletion also increased kynurenine uptake by increasing SLC7A5 (LAT1) expression in a GCN2-dependent manner. Tryptophan deprivation potentiated Treg differentiation from naïve CD4⁺ T cells isolated from mice bearing an AHR allele of weak affinity similar to the human AHR.

CONCLUSIONS

Tryptophan deprivation sensitizes the AHR pathway by inducing AHR overexpression and increasing cellular kynurenine uptake. As a result, tryptophan catabolites such as kynurenine more potently activate AHR, and Treg differentiation is promoted. Our results propose a molecular explanation for the combined roles of tryptophan deprivation and kynurenine production in mediating IDO1/TDO-induced immune suppression.

Multiomics and spatial mapping characterizes human CD8+ T cell states in cancer

Clinically relevant immunological biomarkers that discriminate between diverse hypofunctional states of tumor-associated CD8⁺ T cells remain disputed. Using multiomics analysis of CD8⁺ T cell features across multiple patient cohorts and tumor types, we identified tumor niche-dependent exhausted and other types of hypofunctional CD8⁺ T cell states. CD8⁺ T cells in "supportive" niches, like melanoma or lung cancer, exhibited features of tumor reactivity-driven exhaustion (CD8⁺ T_EX). These included a proficient effector memory phenotype, an expanded T cell receptor (TCR) repertoire linked to effector exhaustion signaling, and a cancer-relevant T cell-activating immunopeptidome composed of largely shared cancer antigens or neoantigens. In contrast, "nonsupportive" niches, like glioblastoma, were enriched for features of hypofunctionality distinct from canonical exhaustion. This included immature or insufficiently activated T cell states, high wound healing signatures, nonexpanded TCR repertoires linked to anti-inflammatory signaling, high T cell-recognizable self-epitopes, and an antiproliferative state linked to stress or prodeath responses. In situ spatial mapping of glioblastoma highlighted the prevalence of dysfunctional CD4⁺:CD8⁺ T cell interactions, whereas ex vivo single-cell secretome mapping of glioblastoma CD8⁺ T cells confirmed negligible effector functionality and a promyeloid, wound healing-like chemokine profile. Within immuno-oncology clinical trials, anti-programmed cell death protein 1 (PD-1) immunotherapy facilitated glioblastoma's tolerogenic disparities, whereas dendritic cell (DC) vaccines partly corrected them. Accordingly, recipients of a DC vaccine for glioblastoma had high effector memory CD8⁺ T cells and evidence of antigen-specific immunity. Collectively, we provide an atlas for assessing different CD8⁺ T cell hypofunctional states in immunogenic versus nonimmunogenic cancers.

Trial watch: Dendritic cell (DC)-based immunotherapy for cancer

Dendritic cell (DC)-based vaccination for cancer treatment has seen considerable development over recent decades. However, this field is currently in a state of flux toward niche-applications, owing to recent paradigm-shifts in immuno-oncology mobilized by T cell-targeting immunotherapies. DC vaccines are typically generated using autologous (patient-derived) DCs exposed to tumor-associated or -specific antigens (TAAs or TSAs), in the presence of immunostimulatory molecules to induce DC maturation, followed by reinfusion into patients. Accordingly, DC vaccines can induce TAA/TSA-specific CD8⁺/CD4⁺ T cell responses. Yet, DC vaccination still shows suboptimal anti-tumor efficacy in the clinic. Extensive efforts are ongoing to improve the immunogenicity and efficacy of DC vaccines, often by employing combinatorial chemo-immunotherapy regimens. In this Trial Watch, we summarize the recent preclinical and clinical developments in this field and discuss the ongoing trends and future perspectives of DC-based immunotherapy for oncological indications.

Abstract 1379: Discovery of tumor-associated, immunogenic peptides presented in a patient-derived, mutant calreticulin-driven myeloproliferative neoplasm cell line

Background: Myeloproliferative neoplasms (MPNs) are a type of chronic blood cancers. A subset are driven by frameshift mutations in the calreticulin gene (CALR). The CALR protein is a lectin chaperone and primarily assists in glycoprotein folding in the endoplasmic reticulum. It is also part of the peptide loading complex, a set of proteins involved in antigen processing and peptide presentation by major histocompatibility complex class I (MHC-I) molecules. All MPN-associated frameshift mutations in CALR lead to a similar change in its C-terminal, altering its protein-protein interactions. Examples include impaired chaperone activity of mutant CALR that impacts protein homeostasis and compromised activity of the MHC-I peptide loading complex. We hypothesized that these impaired functions will alter the repertoire of peptides presented by MHC-I in mutant CALR expressing cells, which we characterized in a patient-derived, mutant-CALR-driven myeloid tumor cell line, Marimo.

Methods: We carried out an immunopeptidome analysis on Marimo cell line derivates expressing either wildtype or mutant CALR to identify potential MPN-associated peptides presented by MHC-I, and validated their immunogenicity in human samples. MHC-I-peptide complexes were immunoprecipitated from these Marimo cell line derivates, followed by peptide elution and analysis by mass spectrometry. A subset of peptides that were found in the mass spectrometry analysis of mutant CALR cells but not wildtype CALR cells were tested for their immunogenicity by assaying the in vitro T cell responses they evoked. Naïve T cells from healthy donors, carrying at least one MHC-I allele that matched the Marimo cells, were primed and expanded with the selected peptides and peptide-specific T cell responses were measured by intracellular staining detecting effector cytokines IFN-γ and TNF-α.

Results:

1. The expression of mutant CALR disrupts the function of the peptide loading complex, leading to reduced recruitment of MHC-I and reducing its cell surface expression by almost 40% compared to cells expressing wildtype CALR.

2. Peptide repertoire of MHC-I is significantly altered by the expression of mutant CALR. Immunopeptidomics revealed a qualitative and quantitative difference in peptides presented by cells expressing wildtype CALR vs mutant CALR

3. The immunogenicity of a curated list of 24 peptides was tested in healthy donors, and at least 6 peptides activated T cell responses. Similar testing in MPN patients is planned.

Conclusions: This is the first study to investigate the impact of mutant CALR on the repertoire of peptides presented by MHC-I in myeloid cells. We discovered tumor-associated, immunogenic peptides that may serve as potential targets of immunotherapy allowing the specific elimination of mutant CALR-expressing cells, which has been a challenge in the treatment of MPNs.

Citation Format: Najla Arshad, Nathalie Vigneron, Cansu Cimen Bozkus, Vincent Stroobant, Stefan Naulaerts, Nina Bhardwaj, Benoît Van den Eynde, Peter Cresswell. Discovery of tumor-associated, immunogenic peptides presented in a patient-derived, mutant calreticulin-driven myeloproliferative neoplasm cell line [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 1379.

New Insights into the Mechanisms of Proteasome-Mediated Peptide Splicing Learned from Comparing Splicing Efficiency by Different Proteasome Subtypes

By tying peptide fragments originally distant in parental proteins, the proteasome can generate spliced peptides that are recognized by CTL. This occurs by transpeptidation involving a peptide-acyl-enzyme intermediate and another peptide fragment present in the catalytic chamber. Four main subtypes of proteasomes exist: the standard proteasome (SP), the immunoproteasome, and intermediate proteasomes β1-β2-β5i (single intermediate proteasome) and β1i-β2-β5i (double intermediate proteasome). In this study, we use a tandem mass tag-quantification approach to study the production of six spliced human antigenic peptides by the four proteasome subtypes. Peptides fibroblast growth factor-5_{172-176/217-220,} tyrosinase_{368-373/336-340}, and gp100_40-42/47-52 are better produced by the SP than the other proteasome subtypes. The peptides SP110_{296-301/286-289}, gp100_{195-202/191or192}, and gp100_47-52/40-42 are better produced by the immunoproteasome and double intermediate proteasome. The current model of proteasome-catalyzed peptide splicing suggests that the production of a spliced peptide depends on the abundance of the peptide splicing partners. Surprisingly, we found that despite the fact that reciprocal peptides RTK_QLYPEW (gp100_40-42/47-52) and QLYPEW_RTK (gp100_47-52/40-42) are composed of identical splicing partners, their production varies differently according to the proteasome subtype. These differences were maintained after in vitro digestions involving identical amounts of the splicing fragments. Our results indicate that the amount of splicing partner is not the only factor driving peptide splicing and suggest that peptide splicing efficiency also relies on other factors, such as the affinity of the C-terminal splice reactant for the primed binding site of the catalytic subunit.

T-cell mediated targeted delivery of anti-PD-L1 nanobody overcomes poor antibody penetration and improves PD-L1 blocking at the tumor site

Monoclonal antibodies blocking immune checkpoints such as PD-L1 have yielded strong clinical benefits in many cancer types. Still, the current limitations are the lack of clinical response in a majority of patients and the development of immune-related adverse events in some. As an alternative to PD-L1-specific antibody injection, we have developed an approach based on the engineering of tumor-targeting T cells to deliver intratumorally an anti-PD-L1 nanobody. In the MC38-OVA model, our strategy enhanced tumor control as compared to injection of PD-L1-specific antibody combined with adoptive transfer of tumor-targeting T cells. As a possible explanation for this, we demonstrated that PD-L1-specific antibody massively occupied PD-L1 in the periphery but failed to penetrate to PD-L1-expressing cells at the tumor site. In sharp contrast, locally delivered anti-PD-L1 nanobody improved PD-L1 blocking at the tumor site while avoiding systemic exposure. Our approach appears promising to overcome the limitations of immunotherapy based on PD-L1-specific antibody treatment.

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.

Concise Polygenic Models for Cancer-Specific Identification of Drug-Sensitive Tumors from Their Multi-Omics Profiles

In silico models to predict which tumors will respond to a given drug are necessary for Precision Oncology. However, predictive models are only available for a handful of cases (each case being a given drug acting on tumors of a specific cancer type). A way to generate predictive models for the remaining cases is with suitable machine learning algorithms that are yet to be applied to existing in vitro pharmacogenomics datasets. Here, we apply XGBoost integrated with a stringent feature selection approach, which is an algorithm that is advantageous for these high-dimensional problems. Thus, we identified and validated 118 predictive models for 62 drugs across five cancer types by exploiting four molecular profiles (sequence mutations, copy-number alterations, gene expression, and DNA methylation). Predictive models were found in each cancer type and with every molecular profile. On average, no omics profile or cancer type obtained models with higher predictive accuracy than the rest. However, within a given cancer type, some molecular profiles were overrepresented among predictive models. For instance, CNA profiles were predictive in breast invasive carcinoma (BRCA) cell lines, but not in small cell lung cancer (SCLC) cell lines where gene expression (GEX) and DNA methylation profiles were the most predictive. Lastly, we identified the best XGBoost model per cancer type and analyzed their selected features. For each model, some of the genes in the selected list had already been found to be individually linked to the response to that drug, providing additional evidence of the usefulness of these models and the merits of the feature selection scheme.

Osmotic stress inhibits leaf growth of Arabidopsis thaliana by enhancing ARF-mediated auxin responses

We investigated the interaction between osmotic stress and auxin signaling in leaf growth regulation. Therefore, we grew Arabidopsis thaliana seedlings on agar media supplemented with mannitol to impose osmotic stress and 1-naphthaleneacetic acid (NAA), a synthetic auxin. We performed kinematic analysis and flow-cytometry to quantify the effects on cell division and expansion in the first leaf pair, determined the effects on auxin homeostasis and response (DR5::β-glucuronidase), performed a next-generation sequencing transcriptome analysis and investigated the response of auxin-related mutants. Mannitol inhibited cell division and expansion. NAA increased the effect of mannitol on cell division, but ameliorated its effect on expansion. In proliferating cells, NAA and mannitol increased free IAA concentrations at the cost of conjugated IAA and stimulated DR5 promotor activity. Transcriptome analysis shows a large overlap between NAA and osmotic stress-induced changes, including upregulation of auxin synthesis, conjugation, transport and TRANSPORT INHIBITOR RESPONSE1 (TIR1) and AUXIN RESPONSE FACTOR (ARF) response genes, but downregulation of Aux/IAA response inhibitors. Consistently, arf7/19 double mutant lack the growth response to auxin and show a significantly reduced sensitivity to osmotic stress. Our results show that osmotic stress inhibits cell division during leaf growth of A. thaliana at least partly by inducing the auxin transcriptional response.

Predicting Synergism of Cancer Drug Combinations Using NCI-ALMANAC Data

Drug combinations are of great interest for cancer treatment. Unfortunately, the discovery of synergistic combinations by purely experimental means is only feasible on small sets of drugs. In silico modeling methods can substantially widen this search by providing tools able to predict which of all possible combinations in a large compound library are synergistic. Here we investigate to which extent drug combination synergy can be predicted by exploiting the largest available dataset to date (NCI-ALMANAC, with over 290,000 synergy determinations). Each cell line is modeled using primarily two machine learning techniques, Random Forest (RF) and Extreme Gradient Boosting (XGBoost), on the datasets provided by NCI-ALMANAC. This large-scale predictive modeling study comprises more than 5,000 pair-wise drug combinations, 60 cell lines, 4 types of models, and 5 types of chemical features. The application of a powerful, yet uncommonly used, RF-specific technique for reliability prediction is also investigated. The evaluation of these models shows that it is possible to predict the synergy of unseen drug combinations with high accuracy (Pearson correlations between 0.43 and 0.86 depending on the considered cell line, with XGBoost providing slightly better predictions than RF). We have also found that restricting to the most reliable synergy predictions results in at least 2-fold error decrease with respect to employing the best learning algorithm without any reliability estimation. Alkylating agents, tyrosine kinase inhibitors and topoisomerase inhibitors are the drugs whose synergy with other partner drugs are better predicted by the models. Despite its leading size, NCI-ALMANAC comprises an extremely small part of all conceivable combinations. Given their accuracy and reliability estimation, the developed models should drastically reduce the number of required in vitro tests by predicting in silico which of the considered combinations are likely to be synergistic.

Precision and recall oncology: combining multiple gene mutations for improved identification of drug-sensitive tumours

Cancer drug therapies are only effective in a small proportion of patients. To make things worse, our ability to identify these responsive patients before administering a treatment is generally very limited. The recent arrival of large-scale pharmacogenomic data sets, which measure the sensitivity of molecularly profiled cancer cell lines to a panel of drugs, has boosted research on the discovery of drug sensitivity markers. However, no systematic comparison of widely-used single-gene markers with multi-gene machine-learning markers exploiting genomic data has been so far conducted. We therefore assessed the performance offered by these two types of models in discriminating between sensitive and resistant cell lines to a given drug. This was carried out for each of 127 considered drugs using genomic data characterising the cell lines. We found that the proportion of cell lines predicted to be sensitive that are actually sensitive (precision) varies strongly with the drug and type of model used. Furthermore, the proportion of sensitive cell lines that are correctly predicted as sensitive (recall) of the best single-gene marker was lower than that of the multi-gene marker in 118 of the 127 tested drugs. We conclude that single-gene markers are only able to identify those drug-sensitive cell lines with the considered actionable mutation, unlike multi-gene markers that can in principle combine multiple gene mutations to identify additional sensitive cell lines. We also found that cell line sensitivities to some drugs (e.g. Temsirolimus, 17-AAG or Methotrexate) are better predicted by these machine-learning models.

Predicting the Reliability of Drug-target Interaction Predictions with Maximum Coverage of Target Space

Many computational methods to predict the macromolecular targets of small organic molecules have been presented to date. Despite progress, target prediction methods still have important limitations. For example, the most accurate methods implicitly restrict their predictions to a relatively small number of targets, are not systematically validated on drugs (whose targets are harder to predict than those of non-drug molecules) and often lack a reliability score associated with each predicted target. Here we present a systematic validation of ligand-centric target prediction methods on a set of clinical drugs. These methods exploit a knowledge-base covering 887,435 known ligand-target associations between 504,755 molecules and 4,167 targets. Based on this dataset, we provide a new estimate of the polypharmacology of drugs, which on average have 11.5 targets below IC₅₀ 10 µM. The average performance achieved across clinical drugs is remarkable (0.348 precision and 0.423 recall, with large drug-dependent variability), especially given the unusually large coverage of the target space. Furthermore, we show how a sparse ligand-target bioactivity matrix to retrospectively validate target prediction methods could underestimate prospective performance. Lastly, we present and validate a first-in-kind score capable of accurately predicting the reliability of target predictions.

Practical Approaches for Mining Frequent Patterns in Molecular Datasets

Pattern detection is an inherent task in the analysis and interpretation of complex and continuously accumulating biological data. Numerous itemset mining algorithms have been developed in the last decade to efficiently detect specific pattern classes in data. Although many of these have proven their value for addressing bioinformatics problems, several factors still slow down promising algorithms from gaining popularity in the life science community. Many of these issues stem from the low user-friendliness of these tools and the complexity of their output, which is often large, static, and consequently hard to interpret. Here, we apply three software implementations on common bioinformatics problems and illustrate some of the advantages and disadvantages of each, as well as inherent pitfalls of biological data mining. Frequent itemset mining exists in many different flavors, and users should decide their software choice based on their research question, programming proficiency, and added value of extra features.

Bioinformatics approaches for the functional interpretation of protein lists: from ontology term enrichment to network analysis

The main result of a great deal of the published proteomics studies is a list of identified proteins, which then needs to be interpreted in relation to the research question and existing knowledge. In the early days of proteomics this interpretation was only based on expert insights, acquired by digesting a large amount of relevant literature. With the growing size and complexity of the experimental datasets, many computational techniques, databases, and tools have claimed a central role in this task. In this review we discuss commonly and less commonly used methods to functionally interpret experimental proteome lists and compare them with available knowledge. We first address several functional analysis and enrichment techniques based on ontologies and literature. Then we outline how various types of network and pathway information can be used. While the problem of functional interpretation of proteome data is to an extent equivalent to the interpretation of transcriptome or other ''omics'' data, this paper addresses some of the specific challenges and solutions of the proteomics field.

Ectopic microRNA-150-5p transcription sensitizes glucocorticoid therapy response in MM1S multiple myeloma cells but fails to overcome hormone therapy resistance in MM1R cells

Glucocorticoids (GCs) selectively trigger cell death in the multiple myeloma cell line MM1S which express NR3C1/Glucocorticoid Receptor (GR) protein, but fail to kill MM1R cells which lack GR protein. Given recent demonstrations of altered microRNA profiles in a diverse range of haematological malignancies and drug resistance, we characterized GC inducible mRNA and microRNA transcription profiles in GC sensitive MM1S as compared to GC resistant MM1R cells. Transcriptome analysis revealed that GCs regulate expression of multiple genes involved in cell cycle control, cell organization, cell death and immunological disease in MM1S cells, which remain unaffected in MM1R cells. With respect to microRNAs, mir-150-5p was identified as the most time persistent GC regulated microRNA, out of 5 QPCR validated microRNAs (mir-26b, mir-125a-5p, mir-146-5p, mir-150-5p, and mir-184), which are GC inducible in MM1S but not in MM1R cells. Functional studies further revealed that ectopic transfection of a synthetic mir-150-5p mimics GR dependent gene expression changes involved in cell death and cell proliferation pathways. Remarkably, despite the gene expression changes observed, overexpression of mir-150-5p in absence of GCs did not trigger significant cytotoxicity in MM1S or MM1R cells. This suggests the requirement of additional steps in GC induced cell death, which can not be mimicked by mir-150-5p overexpression alone. Interestingly, a combination of mir-150-5p transfection with low doses GC in MM1S cells was found to sensitize therapy response, whereas opposite effects could be observed with a mir-150-5p specific antagomir. Although mir-150-5p overexpression did not substantially change GR expression levels, it was found that mir-150-5p evokes GR specific effects through indirect mRNA regulation of GR interacting transcription factors and hormone receptors, GR chaperones, as well as various effectors of unfolded protein stress and chemokine signalling. Altogether GC-inducible mir-150-5p adds another level of regulation to GC specific therapeutic responses in multiple myeloma.

A primer to frequent itemset mining for bioinformatics

Over the past two decades, pattern mining techniques have become an integral part of many bioinformatics solutions. Frequent itemset mining is a popular group of pattern mining techniques designed to identify elements that frequently co-occur. An archetypical example is the identification of products that often end up together in the same shopping basket in supermarket transactions. A number of algorithms have been developed to address variations of this computationally non-trivial problem. Frequent itemset mining techniques are able to efficiently capture the characteristics of (complex) data and succinctly summarize it. Owing to these and other interesting properties, these techniques have proven their value in biological data analysis. Nevertheless, information about the bioinformatics applications of these techniques remains scattered. In this primer, we introduce frequent itemset mining and their derived association rules for life scientists. We give an overview of various algorithms, and illustrate how they can be used in several real-life bioinformatics application domains. We end with a discussion of the future potential and open challenges for frequent itemset mining in the life sciences.