AKT and EZH2 inhibitors kill TNBCs by hijacking mechanisms of involution.

Triple-negative breast cancer (TNBC) is the most aggressive breast cancer subtype and has the highest rate of recurrence1. The predominant standard of care for advanced TNBC is systemic chemotherapy with or without immunotherapy; however, responses are typically short lived1,2. Thus, there is an urgent need to develop more effective treatments. Components of the PI3K pathway represent plausible therapeutic targets; more than 70% of TNBCs have alterations in PIK3CA, AKT1 or PTEN3-6. However, in contrast to hormone-receptor-positive tumours, it is still unclear whether or how triple-negative disease will respond to PI3K pathway inhibitors7. Here we describe a promising AKT-inhibitor-based therapeutic combination for TNBC. Specifically, we show that AKT inhibitors synergize with agents that suppress the histone methyltransferase EZH2 and promote robust tumour regression in multiple TNBC models in vivo. AKT and EZH2 inhibitors exert these effects by first cooperatively driving basal-like TNBC cells into a more differentiated, luminal-like state, which cannot be effectively induced by either agent alone. Once TNBCs are differentiated, these agents kill them by hijacking signals that normally drive mammary gland involution. Using a machine learning approach, we developed a classifier that can be used to predict sensitivity. Together, these findings identify a promising therapeutic strategy for this highly aggressive tumour type and illustrate how deregulated epigenetic enzymes can insulate tumours from oncogenic vulnerabilities. These studies also reveal how developmental tissue-specific cell death pathways may be co-opted for therapeutic benefit.

Multifidelity Neural Network Formulations for Prediction of Reactive Molecular Potential Energy Surfaces.

This paper focuses on the development of multifidelity modeling approaches using neural network surrogates, where training data arising from multiple model forms and resolutions are integrated to predict high-fidelity response quantities of interest at lower cost. We focus on the context of quantum chemistry and the integration of information from multiple levels of theory. Important foundations include the use of symmetry function-based atomic energy vector constructions as feature vectors for representing structures across families of molecules and single-fidelity neural network training capabilities that learn the relationships needed to map feature vectors to potential energy predictions. These foundations are embedded within several multifidelity topologies that decompose the high-fidelity mapping into model-based components, including sequential formulations that admit a general nonlinear mapping across fidelities and discrepancy-based formulations that presume an additive decomposition. Methodologies are first explored and demonstrated on a pair of simple analytical test problems and then deployed for potential energy prediction for C5H5 using B2PLYP-D3/6-311++G(d,p) for high-fidelity simulation data and Hartree-Fock 6-31G for low-fidelity data. For the common case of limited access to high-fidelity data, our computational results demonstrate that multifidelity neural network potential energy surface constructions achieve roughly an order of magnitude improvement, either in terms of test error reduction for equivalent total simulation cost or reduction in total cost for equivalent error.

Automated Reaction Kinetics of Gas-Phase Organic Species over Multiwell Potential Energy Surfaces.

Automation of rate-coefficient calculations for gas-phase organic species became possible in recent years and has transformed how we explore these complicated systems computationally. Kinetics workflow tools bring rigor and speed and eliminate a large fraction of manual labor and related error sources. In this paper we give an overview of this quickly evolving field and illustrate, through five detailed examples, the capabilities of our own automated tool, KinBot. We bring examples from combustion and atmospheric chemistry of C-, H-, O-, and N-atom-containing species that are relevant to molecular weight growth and autoxidation processes. The examples shed light on the capabilities of automation and also highlight particular challenges associated with the various chemical systems that need to be addressed in future work.

Quantification of DNA/SWCNT Solvation Differences by Aqueous Two-Phase Separation.

Single-walled carbon nanotubes (SWCNTs) coated with single-stranded DNA can be effectively separated into various chiralities using an aqueous two-phase (ATP) system. Partitioning is driven by small differences in the dissolution characteristics of the hybrid between the two phases. Thus, in addition to being a separation technique, the ATP system potentially also offers a way to quantify and rank the dissolution properties of the solute (here the DNA/SWCNT hybrids), such as the solvation free energy and solubility. In this study, we propose two different approaches to quantitatively analyze the ATP partitioning of DNA/SWCNT hybrids. First, we present a model that extracts the relative solvation free energy of various DNA/SWCNT hybrids by using an expansion relative to a standard state. Second, we extract a solubility parameter by analyzing the partitioning of hybrids in the ATP system. The two approaches are found to be consistent, providing some confidence in each as a method of quantifying differences in the solubility of various DNA/SWCNT hybrids.

The molecular consequences of androgen activity in the human breast.

Estrogen and progesterone have been extensively studied in the mammary gland, but the molecular effects of androgen remain largely unexplored. Transgender men are recorded as female at birth but identify as male and may undergo gender-affirming androgen therapy to align their physical characteristics and gender identity. Here we perform single-cell-resolution transcriptome, chromatin, and spatial profiling of breast tissues from transgender men following androgen therapy. We find canonical androgen receptor gene targets are upregulated in cells expressing the androgen receptor and that paracrine signaling likely drives sex-relevant androgenic effects in other cell types. We also observe involution of the epithelium and a spatial reconfiguration of immune, fibroblast, and vascular cells, and identify a gene regulatory network associated with androgen-induced fat loss. This work elucidates the molecular consequences of androgen activity in the human breast at single-cell resolution.

Machine Learning-Guided Systematic Search of DNA Sequences for Sorting Carbon Nanotubes.

The prerequisite of utilizing DNA in sequence-dependent applications is to search specific sequences. Developing a strategy for efficient DNA sequence screening represents a grand challenge due to the countless possibilities of sequence combination. Herein, relying on sequence-dependent recognition between DNA and single-wall carbon nanotubes (SWCNTs), we demonstrate a method for systematic search of DNA sequences for sorting single-chirality SWCNTs. Different from previously documented empirical search, which has a low efficiency and accuracy, our approach combines machine learning and experimental investigation. The number of resolving sequences and the success rate of finding them are improved from ∼102 to ∼103 and from ∼10% to >90%, respectively. Moreover, the resolving sequence patterns determined from 5-mer and 6-mer short sequences can be extended to sequence search in longer DNA subspaces.

Detection of ovarian cancer via the spectral fingerprinting of quantum-defect-modified carbon nanotubes in serum by machine learning.

Serum biomarkers are often insufficiently sensitive or specific to facilitate cancer screening or diagnostic testing. In ovarian cancer, the few established serum biomarkers are highly specific, yet insufficiently sensitive to detect early-stage disease and to impact the mortality rates of patients with this cancer. Here we show that a 'disease fingerprint' acquired via machine learning from the spectra of near-infrared fluorescence emissions of an array of carbon nanotubes functionalized with quantum defects detects high-grade serous ovarian carcinoma in serum samples from symptomatic individuals with 87% sensitivity at 98% specificity (compared with 84% sensitivity at 98% specificity for the current best clinical screening test, which uses measurements of cancer antigen 125 and transvaginal ultrasonography). We used 269 serum samples to train and validate several machine-learning classifiers for the discrimination of patients with ovarian cancer from those with other diseases and from healthy individuals. The predictive values of the best classifier could not be attained via known protein biomarkers, suggesting that the array of nanotube sensors responds to unidentified serum biomarkers.

A perception-based nanosensor platform to detect cancer biomarkers.

Conventional molecular recognition elements, such as antibodies, present issues for developing biomolecular assays for use in certain technologies, such as implantable devices. Additionally, antibody development and use, especially for highly multiplexed applications, can be slow and costly. We developed a perception-based platform based on an optical nanosensor array that leverages machine learning algorithms to detect multiple protein biomarkers in biofluids. We demonstrated this platform in gynecologic cancers, often diagnosed at advanced stages, leading to low survival rates. We investigated the detection of protein biomarkers in uterine lavage samples, which are enriched with certain cancer markers compared to blood. We found that the method enables the simultaneous detection of multiple biomarkers in patient samples, with F1-scores of ~0.95 in uterine lavage samples from patients with cancer. This work demonstrates the potential of perception-based systems for the development of multiplexed sensors of disease biomarkers without the need for specific molecular recognition elements.