Background: The genetic bases of many complex phenotypes are still largely unknown, mostly due to the polygenic nature of the traits and the small effect of each associated mutation. An alternative approach to classic association studies to determining such genetic bases is an evolutionary framework. As sites targeted by natural selection are likely to harbor important functionalities for the carrier, the identification of selection signatures in the genome has the potential to unveil the genetic mechanisms underpinning human phenotypes. Popular methods of detecting such signals rely on compressing genomic information into summary statistics, resulting in the loss of information. Furthermore, few methods are able to quantify the strength of selection. Here we explored the use of deep learning in evolutionary biology and implemented a program, called , to apply convolutional neural networks on population genomic data for the detection and quantification of natural selection.Results: enables genomic information from multiple individuals to be represented as abstract images. Each image is created by stacking aligned genomic data and encoding distinct alleles into separate colors. To detect and quantify signatures of positive selection, implements a convolutional neural network which is trained using simulations. We show how the method implemented in can be affected by data manipulation and learning strategies. In particular, we show how sorting images by row and column leads to accurate predictions. We also demonstrate how the misspecification of the correct demographic model for producing training data can influence the quantification of positive selection. We finally illustrate an approach to estimate the selection coefficient, a continuous variable, using multiclass classification techniques.Conclusions: While the use of deep learning in evolutionary genomics is in its infancy, here we demonstrated its potential to detect informative patterns from large-scale genomic data. We implemented methods to process genomic data for deep learning in a user-friendly program called . The joint inference of the evolutionary history of mutations and their functional impact will facilitate mapping studies and provide novel insights into the molecular mechanisms associated with human phenotypes.

Elucidating the molecular basis of tumor metastasis is pivotal for eradicating cancer-related mortality. Triple-negative breast cancer (TNBC) encompasses a class of aggressive tumors characterized by high rates of recurrence and metastasis, as well as poor overall survival. Here, we find that the promyelocytic leukemia protein PML exerts a prometastatic function in TNBC that can be targeted by arsenic trioxide. We found that, in TNBC patients, constitutive HIF1A activity induces high expression of PML, along with a number of HIF1A target genes that promote metastasis at multiple levels. Intriguingly, PML controls the expression of these genes by binding to their regulatory regions along with HIF1A. This mechanism is specific to TNBC cells and does not occur in other subtypes of breast cancer where PML and prometastatic HIF1A target genes are underexpressed. As a consequence, PML promotes cell migration, invasion, and metastasis in TNBC cell and mouse models. Notably, pharmacological inhibition of PML with arsenic trioxide, a PML-degrading agent used to treat promyelocytic leukemia patients, delays tumor growth, impairs TNBC metastasis, and cooperates with chemotherapy by preventing metastatic dissemination. In conclusion, we report identification of a prometastatic pathway in TNBC and suggest clinical development toward the use of arsenic trioxide for TNBC patients.

All-trans retinoic acid (ATRA), a recognized differentiating agent, has significant potential in the personalized/stratified treatment of breast cancer. The present study reports on the molecular mechanisms underlying the anti-tumor activity of ATRA in breast cancer. The work is based on transcriptomic experiments performed on ATRA-treated breast cancer cell-lines, short-term tissue cultures of patient-derived mammary-tumors and a xenograft model. ATRA upregulates gene networks involved in interferon-responses, immune-modulation and antigen-presentation in retinoid-sensitive cells and tumors characterized by poor immunogenicity. ATRA-dependent upregulation of these gene networks is caused by a viral mimicry process, involving the activation of endogenous retroviruses. ATRA induces a non-canonical type of viral mimicry, which results in increased expression of the IRF1 (Interferon Responsive Factor 1) transcription factor and the DTX3L (Deltex-E3-Ubiquitin-Ligase-3L) downstream effector. Functional knockdown studies indicate that IRF1 and DTX3L are part of a negative feedback loop controlling ATRA-dependent growth inhibition of breast cancer cells. The study is of relevance from a clinical/therapeutic perspective. In fact, ATRA stimulates processes controlling the sensitivity to immuno-modulatory drugs, such as immune-checkpoint-inhibitors. This suggests that ATRA and immunotherapeutic agents represent rational combinations for the personalized treatment of breast cancer. Remarkably, ATRA-sensitivity seems to be relatively high in immune-cold mammary tumors, which are generally resistant to immunotherapy.

In chronic lymphocytic leukemia (CLL), the non-hematopoietic stromal microenvironment plays a critical role in promoting tumor cell recruitment, activation, survival, and expansion. However, the nature of the stromal cells and molecular pathways involved remain largely unknown. Here, we demonstrate that leukemic B lymphocytes induce the activation of retinoid acid synthesis and signaling in the microenvironment. Inhibition of RA-signaling in stromal cells causes deregulation of genes associated with adhesion, tissue organization and chemokine secretion including the B-cell chemokine CXCL13. Notably, reducing retinoic acid precursors from the diet or inhibiting RA-signaling through retinoid-antagonist therapy prolong survival by preventing dissemination of leukemia cells into lymphoid tissues. Furthermore, mouse and human leukemia cells could be distinguished from normal B-cells by their increased expression ofRarγ2andRXRα,respectively. These findings establish a role for retinoids in murine CLL pathogenesis, and provide new therapeutic strategies to target the microenvironment and to control disease progression.

Background: Acute promyelocytic leukemia (APL) is a sub-type of acute myeloid leukemia (AML) characterized by a block of myeloid differentiation at the promyelocytic stage and the predominant t(15:17) chromosomal translocation. We have previously determined that cells from APL patients show increased expression of genes regulated by hypoxia-inducible transcription factors (HIFs) compared to normal promyelocytes. HIFs regulate crucial aspects of solid tumor progression and are currently being implicated in leukemogenesis. Methods: To investigate the contribution of hypoxia-related signaling in APL compared to other AML sub-types, we reverse engineered a transcriptional network from gene expression profiles of AML patients’ samples, starting from a list of direct target genes of HIF-1. A HIF-1-dependent subnetwork of genes specifically dysregulated in APL was derived from the comparison between APL and other AMLs. Results: Interestingly, this subnetwork shows a unique involvement of genes related to extracellular matrix interaction and cell migration, with decreased expression of genes involved in cell adhesion and increased expression of genes implicated in motility and invasion, thus unveiling the presence of characteristics of epithelial-mesenchymal transition (EMT). We observed that the genes of this subnetwork, whose dysregulation shows a peculiar pattern across different AML sub-types, distinguish malignant from normal promyelocytes, thus ruling out dependence on a myeloid developmental stage. Also, expression of these genes is reversed upon treatment of APL-derived NB4 cells with all-trans retinoic acid and cell differentiation. Conclusions: Our data suggest that pathways related to EMT-like processes can be implicated also in hematological malignancies besides solid tumors, and can identify specific AML sub-types.

Lung neuroendocrine neoplasms (lung NENs) are categorised by morphology, defining a classification sometimes unable to reflect ultimate clinical outcome. Subjectivity and poor reproducibility characterise diagnosis and prognosis assessment of all NENs. Here, we propose a machine learning framework for tumour prognosis assessment based on a quantitative, automated and repeatable evaluation of the spatial distribution of cells immunohistochemically positive for the proliferation marker Ki-67, performed on the entire extent of high-resolution whole slide images. Combining features from the fields of graph theory, fractality analysis, stochastic geometry and information theory, we describe the topology of replicating cells and predict prognosis in a histology-independent way. We demonstrate how our approach outperforms the well-recognised prognostic role of Ki-67 Labelling Index on a multi-centre dataset comprising the most controversial lung NENs. Moreover, we show that our system identifies arrangement patterns in the cells positive for Ki-67 that appear independently of tumour subtyping. Strikingly, the subset of these features whose presence is also independent of the value of the Labelling Index and the density of Ki-67-positive cells prove to be especially relevant in discerning prognostic classes. These findings disclose a possible path for the future of grading and classification of NENs.