Deep Morphology Learning Enhances Ex Vivo Drug Profiling-Based Precision Medicine.

Drug testing in patient biopsy-derived cells can identify potent treatments for patients suffering from relapsed or refractory hematologic cancers. Here we investigate the use of weakly supervised deep learning on cell morphologies (DML) to complement diagnostic marker-based identification of malignant and nonmalignant cells in drug testing. Across 390 biopsies from 289 patients with diverse blood cancers, DML-based drug responses show improved reproducibility and clustering of drugs with the same mode of action. DML does so by adapting to batch effects and by autonomously recognizing disease-associated cell morphologies. In a post hoc analysis of 66 patients, DML-recommended treatments led to improved progression-free survival compared with marker-based recommendations and physician's choice-based treatments. Treatments recommended by both immunofluorescence and DML doubled the fraction of patients achieving exceptional clinical responses. Thus, DML-enhanced ex vivo drug screening is a promising tool in the identification of effective personalized treatments. SIGNIFICANCE: We have recently demonstrated that image-based drug screening in patient samples identifies effective treatment options for patients with advanced blood cancers. Here we show that using deep learning to identify malignant and nonmalignant cells by morphology improves such screens. The presented workflow is robust, automatable, and compatible with clinical routine. This article is highlighted in the In This Issue feature, p. 476.

Functional Precision Medicine Provides Clinical Benefit in Advanced Aggressive Hematologic Cancers and Identifies Exceptional Responders.

Personalized medicine aims to match the right drug with the right patient by using specific features of the individual patient's tumor. However, current strategies of personalized therapy matching provide treatment opportunities for less than 10% of patients with cancer. A promising method may be drug profiling of patient biopsy specimens with single-cell resolution to directly quantify drug effects. We prospectively tested an image-based single-cell functional precision medicine (scFPM) approach to guide treatments in 143 patients with advanced aggressive hematologic cancers. Fifty-six patients (39%) were treated according to scFPM results. At a median follow-up of 23.9 months, 30 patients (54%) demonstrated a clinical benefit of more than 1.3-fold enhanced progression-free survival compared with their previous therapy. Twelve patients (40% of responders) experienced exceptional responses lasting three times longer than expected for their respective disease. We conclude that therapy matching by scFPM is clinically feasible and effective in advanced aggressive hematologic cancers. SIGNIFICANCE: This is the first precision medicine trial using a functional assay to instruct n-of-one therapies in oncology. It illustrates that for patients lacking standard therapies, high-content assay-based scFPM can have a significant value in clinical therapy guidance based on functional dependencies of each patient's cancer.See related commentary by Letai, p. 290.This article is highlighted in the In This Issue feature, p. 275.

Modeling chemotherapy-induced stress to identify rational combination therapies in the DNA damage response pathway.

Cells respond to DNA damage by activating complex signaling networks that decide cell fate, promoting not only DNA damage repair and survival but also cell death. We have developed a multiscale computational model that quantitatively links chemotherapy-induced DNA damage response signaling to cell fate. The computational model was trained and calibrated on extensive data from U2OS osteosarcoma cells, including the cell cycle distribution of the initial cell population, signaling data measured by Western blotting, and cell fate data in response to chemotherapy treatment measured by time-lapse microscopy. The resulting mechanistic model predicted the cellular responses to chemotherapy alone and in combination with targeted inhibitors of the DNA damage response pathway, which we confirmed experimentally. Computational models such as the one presented here can be used to understand the molecular basis that defines the complex interplay between cell survival and cell death and to rationally identify chemotherapy-potentiating drug combinations.

Model calibration and uncertainty analysis in signaling networks.

For a long time the biggest challenges in modeling cellular signal transduction networks has been the inference of crucial pathway components and the qualitative description of their interactions. As a result of the emergence of powerful high-throughput experiments, it is now possible to measure data of high temporal and spatial resolution and to analyze signaling dynamics quantitatively. In addition, this increase of high-quality data is the basis for a better understanding of model limitations and their influence on the predictive power of models. We review established approaches in signal transduction network modeling with a focus on ordinary differential equation models as well as related developments in model calibration. As central aspects of the calibration process we discuss possibilities of model adaptation based on data-driven parameter optimization and the concomitant objective of reducing model uncertainties.

Robustness of DNA repair through collective rate control.

DNA repair and other chromatin-associated processes are carried out by enzymatic macromolecular complexes that assemble at specific sites on the chromatin fiber. How the rate of these molecular machineries is regulated by their constituent parts is poorly understood. Here we quantify nucleotide-excision DNA repair in mammalian cells and find that, despite the pathways' molecular complexity, repair effectively obeys slow first-order kinetics. Theoretical analysis and data-based modeling indicate that these kinetics are not due to a singular rate-limiting step. Rather, first-order kinetics emerge from the interplay of rapidly and reversibly assembling repair proteins, stochastically distributing DNA lesion repair over a broad time period. Based on this mechanism, the model predicts that the repair proteins collectively control the repair rate. Exploiting natural cell-to-cell variability, we corroborate this prediction for the lesion-recognition factor XPC and the downstream factor XPA. Our findings provide a rationale for the emergence of slow time scales in chromatin-associated processes from fast molecular steps and suggest that collective rate control might be a widespread mode of robust regulation in DNA repair and transcription.

Creating functional engineered variants of the single-module non-ribosomal peptide synthetase IndC by T domain exchange.

Non-ribosomal peptide synthetases (NRPSs) are enzymes that catalyze ribosome-independent production of small peptides, most of which are bioactive. NRPSs act as peptide assembly lines where individual, often interconnected modules each incorporate a specific amino acid into the nascent chain. The modules themselves consist of several domains that function in the activation, modification and condensation of the substrate. NRPSs are evidently modular, yet experimental proof of the ability to engineer desired permutations of domains and modules is still sought. Here, we use a synthetic-biology approach to create a small library of engineered NRPSs, in which the domain responsible for carrying the activated amino acid (T domain) is exchanged with natural or synthetic T domains. As a model system, we employ the single-module NRPS IndC from Photorhabdus luminescens that produces the blue pigment indigoidine. As chassis we use Escherichia coli. We demonstrate that heterologous T domain exchange is possible, even for T domains derived from different organisms. Interestingly, substitution of the native T domain with a synthetic one enhanced indigoidine production. Moreover, we show that selection of appropriate inter-domain linker regions is critical for functionality. Taken together, our results extend the engineering avenues for NRPSs, as they point out the possibility of combining domain sequences coming from different pathways, organisms or from conservation criteria. Moreover, our data suggest that NRPSs can be rationally engineered to control the level of production of the corresponding peptides. This could have important implications for industrial and medical applications.

TIN-a combinatorial compound collection of synthetically feasible multicomponent synthesis products.

The synthetic feasibility of any compound library used for virtual screening is critical to the drug discovery process. TIN, a recursive acronym for 'TIN Is Not commercial', is a virtual combinatorial database enumeration of diversity-orientated multicomponent syntheses (MCR). Using a 'one-pot' synthetic technique, 12 unique small molecule scaffolds were developed, predominantly styrylisoxazoles and bis-acetylenic ketones, with extensive derivatization potential. Importantly, the scaffolds were accessible in a single operation from commercially available sources containing R-groups which were then linked combinatorially. This resulted in a combinatorial database of over 28 million product structures, each of which is synthetically feasible. These structures can be accessed through a free Web-based 2D structure search engine or downloaded in SMILES, MOL2, and SDF formats. Subsets include a 10% diversity subset, a drug-like subset, and a lead-like subset that are also freely available for download and virtual screening ( http://mmg.rcsi.ie:8080/tin ).