Algorithm for cellular reprogramming

Abstract

The day we understand the time evolution of subcellular events at a level of detail comparable to physical systems governed by Newton's laws of motion seems far away. Even so, quantitative approaches to cellular dynamics add to our understanding of cell biology. With data-guided frameworks we can develop better predictions about, and methods for, control over specific biological processes and system-wide cell behavior. Here we describe an approach for optimizing the use of transcription factors (TFs) in cellular reprogramming, based on a device commonly used in optimal control. We construct an approximate model for the natural evolution of a cell-cycle-synchronized population of human fibroblasts, based on data obtained by sampling the expression of 22,083 genes at several time points during the cell cycle. To arrive at a model of moderate complexity, we cluster gene expression based on division of the genome into topologically associating domains (TADs) and then model the dynamics of TAD expression levels. Based on this dynamical model and additional data, such as known TF binding sites and activity, we develop a methodology for identifying the top TF candidates for a specific cellular reprogramming task. Our data-guided methodology identifies a number of TFs previously validated for reprogramming and/or natural differentiation and predicts some potentially useful combinations of TFs. Our findings highlight the immense potential of dynamical models, mathematics, and data-guided methodologies for improving strategies for control over biological processes.

Keywords: cellular reprogramming; control theory; genome architecture; networks; time series data.

Fig. 1.

Overview of TAD dimension reduction. (A) Partitioning the Hi-C matrix based on the Fiedler vector. (B) Cartoon depiction of TAD genomic structure. (C) TAD dimension reduction summary.

Fig. 2.

DGC overview. (A) Summary of control equation variables. (B) Each TAD is a node in a dynamic network. The blue connections represent the edges of the network and are determined from time series fibroblast RNA-seq data. The green plots represent the expression of each TAD changing over time. The red arrows indicate additional regulation imposed by exogenous TFs. (C) A conceptual illustration of the problem: Can we determine TFs to push the cell state from one basin to another?

Fig. 3.

Quantitative measure between cell types and TF scores. (A) $d_0$ values between GTEx tissue types and ESC, myotube, and fibroblast. Tissue types and cell types with black arrows have predicted TFs for reprogramming from fibroblasts shown in B. (B) Table of predicted TFs for a subset of cell and tissue types. Top five TFs for combinations of one to three are shown. Green labeled TFs are highly associated with the differentiation process of the target cell type and/or validated for reprogramming. These TFs are discussed in the main text. (C) Time-dependent scores for selected combinations of three TFs for fibroblast to ESC and fibroblast to “heart - left ventricle.” x axis refers to time of TF addition, and y axis refers to $\mu$.