A single-cell analysis of breast cancer cell lines to study tumour heterogeneity and drug response

Cancer cells within a tumour have heterogeneous phenotypes and exhibit dynamic plasticity. How to evaluate such heterogeneity and its impact on outcome and drug response is still unclear. Here, we transcriptionally profile 35,276 individual cells from 32 breast cancer cell lines to yield a single cell atlas. We find high degree of heterogeneity in the expression of biomarkers. We then train a deconvolution algorithm on the atlas to determine cell line composition from bulk gene expression profiles of tumour biopsies, thus enabling cell line-based patient stratification. Finally, we link results from large-scale in vitro drug screening in cell lines to the single cell data to computationally predict drug responses starting from single-cell profiles. We find that transcriptional heterogeneity enables cells with differential drug sensitivity to co-exist in the same population. Our work provides a framework to determine tumour heterogeneity in terms of cell line composition and drug response.

An experimental approach to identify dynamical models of transcriptional regulation in living cells

We describe an innovative experimental approach, and a proof of principle investigation, for the application of System Identification techniques to derive quantitative dynamical models of transcriptional regulation in living cells. Specifically, we constructed an experimental platform for System Identification based on a microfluidic device, a time-lapse microscope, and a set of automated syringes all controlled by a computer. The platform allows delivering a time-varying concentration of any molecule of interest to the cells trapped in the microfluidics device (input) and real-time monitoring of a fluorescent reporter protein (output) at a high sampling rate. We tested this platform on the GAL1 promoter in the yeast Saccharomyces cerevisiae driving expression of a green fluorescent protein (Gfp) fused to the GAL1 gene. We demonstrated that the System Identification platform enables accurate measurements of the input (sugars concentrations in the medium) and output (Gfp fluorescence intensity) signals, thus making it possible to apply System Identification techniques to obtain a quantitative dynamical model of the promoter. We explored and compared linear and nonlinear model structures in order to select the most appropriate to derive a quantitative model of the promoter dynamics. Our platform can be used to quickly obtain quantitative models of eukaryotic promoters, currently a complex and time-consuming process.

Inference of gene networks from temporal gene expression profiles

Genes interact with each other in complex networks that enable the processing of information and the metabolism of nutrients inside the cell. A novel inference algorithm based on linear ordinary differential equations is proposed. The algorithm can infer the local network of gene-gene interactions surrounding a gene of interest from time-series gene expression profiles. The performance of the algorithm has been tested on in silico simulated gene expression data and on a nine gene subnetwork part of the DNA-damage response pathway (SOS pathway) in the bacteria Escherichia coli. This approach can infer regulatory interactions even when only a small number of measurements is available.

Linear matrix inequalities approach to reconstruction of biological networks

The general problem of reconstructing a biological interaction network from temporal evolution data is tackled via an approach based on dynamical linear systems identification theory. A novel algorithm, based on linear matrix inequalities, is devised to infer the interaction network. This approach allows to directly taking into account, within the optimisation procedure, the a priori available knowledge of the biological system. The effectiveness of the proposed algorithm is statistically validated, by means of numerical tests, demonstrating how the a priori knowledge positively affects the reconstruction performance. A further validation is performed through an in silico biological experiment, exploiting the well-assessed cell-cycle model of fission yeast developed by Novak and Tyson.

A multi-step approach to time series analysis and gene expression clustering

MOTIVATION

The huge growth in gene expression data calls for the implementation of automatic tools for data processing and interpretation.

RESULTS

We present a new and comprehensive machine learning data mining framework consisting in a non-linear PCA neural network for feature extraction, and probabilistic principal surfaces combined with an agglomerative approach based on Negentropy aimed at clustering gene microarray data. The method, which provides a user-friendly visualization interface, can work on noisy data with missing points and represents an automatic procedure to get, with no a priori assumptions, the number of clusters present in the data. Cell-cycle dataset and a detailed analysis confirm the biological nature of the most significant clusters.

AVAILABILITY

The software described here is a subpackage part of the ASTRONEURAL package and is available upon request from the corresponding author.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

The transcriptional landscape of the mammalian genome

This study describes comprehensive polling of transcription start and termination sites and analysis of previously unidentified full-length complementary DNAs derived from the mouse genome. We identify the 5' and 3' boundaries of 181,047 transcripts with extensive variation in transcripts arising from alternative promoter usage, splicing, and polyadenylation. There are 16,247 new mouse protein-coding transcripts, including 5154 encoding previously unidentified proteins. Genomic mapping of the transcriptome reveals transcriptional forests, with overlapping transcription on both strands, separated by deserts in which few transcripts are observed. The data provide a comprehensive platform for the comparative analysis of mammalian transcriptional regulation in differentiation and development.

Dispersion of QT intervals: a measure of dispersion of repolarization or simply a projection effect?

QT interval dispersion may provide little information about repolarization dispersion. Some clinical measurements demonstrate an association between high QT interval dispersion and high morbidity and mortality, but what is being measured is not clear. This study was designed to help resolve this dilemma. We compared the association between different clinical measures of QT interval dispersion and the ECG lead amplitudes derived from a heart vector model of repolarization with no repolarization dispersion whatsoever. We compared our clinical QT interval dispersion data obtained from 25 subjects without cardiac disease with similar data from published studies, and correlated these QT dispersion results with the distribution of lead amplitudes derived from the projection of the heart vector onto the body surface during repolarization. Published results were available for mean relative QT intervals and mean differences from the maximum QT interval. The leads were derived from Uijen and Dower lead vector data. Using the Uijen lead vector data, the correlation between measurements of dispersion and derived lead amplitudes ranged from 0.78 to 0.99 for limb leads, and using the Dower values ranged from 0.81 to 0.94 for the precordial leads. These results show a clear association between the measured QT interval dispersion and the variation in ECG lead amplitudes derived from a simple heart vector model of repolarization with no regional information. Therefore, measured QT dispersion is related mostly to a projection effect and is not a true measure of repolarization dispersion. Our existing interpretation of QT dispersion must be reexamined, and other measurements that provide true repolarization dispersion data investigated.

Computer model for study of cardiac repolarization

INTRODUCTION

We propose a new and simple method to model repolarization in the left ventricle and the corresponding T waves on the surface ECG.

METHODS AND RESULTS

We modeled the cardiac cell action potentials (APs) in the left ventricle (LV) with differences in only the duration of the plateau phase. Using published experimental data on the epicardial and endocardial repolarization sequences, for each point on the left ventricular surface we set a different AP repolarization starting time, determined by the duration of the plateau phase. The surface source model was used to compute potentials on the surface of the torso, generated by repolarization of the LV. Both the torso and the LV had homogeneous and isotropic conductivity. We simulated T waves on the 12-lead ECG and compared our results with measured T waves from five normal subjects. The orientation and shape in each lead were reproduced. In each lead we computed the root mean square error between simulated and measured T waves. The average error across the 12 leads was small, with a mean value of 0.11 mV across all the subjects.

CONCLUSION

Repolarization of the LV can be modeled independently of the depolarization sequence and AP duration gradients. This method is an easy and powerful tool to describe the ECG features of repolarization.