Desired
configuration
Initial configuration
Given an initial condition, target configuration, and the dynamics, are there input(s) that steer the system towards the desired configuration?
Controllability: An engineering perspective
Discrete model
Controllability: A genome perspective
TF Binding data from TRANSFAC and JASPAR to define a B matrix for each TF
The Team
Cellular Reprogramming and Controllability of Complex Systems
Adjacency Matrix from Data
Adjacency Matrix
Intra-chromosomal contacts
Inter-chromosomal
contacts
Hi-C
(1880 x 1880)
Active
Inactive
Mixed
LOC100
TTC29
ZNF827
SLC10A7
PROM1
FGFBP2
FGFBP1
CD38
LSM6
POU4F2
CLOCK
SRD5A3
TMEM165
8
16
24
32
40
48
56
0
0
-1
-0.5
0.5
1
1.5
2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Time (h)
Function (RPKM)
PNAS 2015
Chromatin organization
(static information is shown)
Gene expression over time
human fibroblasts
TADs
4DN Data
Our methods and data for the algorithm
TADs
G1
M
S
G2
Cell Cycle
Time-dependent state-transition matrix
A Matrix
Closest transition matrix to Identity subject to constraints from the data
Natural dimension reduction
G1
M
S
G2
Cell Cycle
Data + Biology = Dimension-Reduced Data
Gene-Level
RNA-seq
TADs
Topological Associated Domains
TAD-Level
RNA-seq
+
=
Gene A
Transcription factor = Protein that binds to DNA and changes local gene expression
Gene A
Gene A
Gene Expression
Gene Expression
Gene Expression
Gene A
Gene A
Gene A
Controllability – Control Input
Gene A
Cannot bind to protein inhibited regions
Gene A
Gene Expression
Gene Expression
Gene Expression
Gene A
Gene A
Gene A
Gene A
Controllability – Control Input
Gene A
Gene A
Gene A
Gene A
expression
Gene A
expression
Activator
Repressor
Transcription factors can be activators or repressors
Controllability – Control Input
Closed or open
Activator or Repressor
From Hi-C
Controllability – Control Input
Controllability – TF Scoring
Controllability – TF Scoring
[1] Lawson, C. L. and R. J. Hanson. Solving Least-Squares Problems. Upper Saddle River, NJ: Prentice Hall. 1974. Chapter 23, p. 161.
An overview of “Algorithm for cellular reprogramming”
2000 functional units inferred
Mathematics + Bioinformatics
(TRANSFAC and FIMO databases)
Dimension reduction using intrinsic delineations in the genome (TADs)
Our algorithm predicted transcription factors
(TF) with known reprogramming capability2
Prediction: any cell to any cell!
TAD: Topologically associating domain, an inherent unit of chromatin organization
TAD1 = g1+g2+g3
TAD2 = g4+g5
TAD3 = g6
+ Gene expression
Initial partioning using Fiedler vector1
Genome-wide contact map (Hi-C)
Rank | TF |
1 | MyoD |
2 | PitX2 |
3 | PKNOX2+ |
4 | Six2 |
Skin to Muscle
Weintraub, 1989
Rank | TF |
1 | NANOG+OCT4+SOX2 |
2-11 | NANOG+OCT4+other |
12 | SOX2+OCT4+MYCN |
Skin to Stem Cell
Yamanaka, 2007