Multiplex analysis of molecules in single cells by image correlation
Disclosed herein are methods and systems for analyzing visual data from multiple rounds of hybridization interactions where the same molecular target is detected by probes with different detectable labels. In particular, disclosed herein are methods and systems for analyzing sequential hybridization images for molecular profiling, where the images are obtained using multiplex fluorescence in situ hybridization (FISH).
1. A method of determining an abundance level of at least two different molecular targets within a defined region in a cell, comprising:
(a) providing a sample comprising the cell;
(b) carrying out a plurality of {1, 2, . . . N} hybridizations on the sample, wherein N is equal or greater than 2, and wherein the plurality of hybridizations interacts with the at least two molecular targets, and wherein each hybridization is performed on the sample by applying a hybridization-specific plurality of probes to the cell, wherein each plurality of probes comprises at least two probes, one probe interacting with at least one of the two different molecular targets, and the other probe interacting with at least the other of the two different molecular targets;
(c) imaging the sample after each hybridization and obtaining an image produced by the interaction between the at least two probes of that hybridization and the at least two molecular targets, wherein the image at least partially constitutes visual data for the hybridization, and wherein at least one image per fluorophore channel per hybridization is taken;
(d) identifying, in the defined region in the cell, an abundance of the at least two molecular targets by using a correlation function, wherein the correlation function spatially correlates visual data from two or more images; and
(e) determining an abundance level of the at least two molecular targets based on the correlation function.
2. The method of claim 1 , wherein the visual data comprise at least two images of the defined region of the cell for each hybridization, and wherein each of the at least two images has predetermined pixel values.
3. The method of claim 1 , wherein the defined region in the cell is the entire cell.
4. The method of claim 1 , wherein the defined region in the cell comprises a portion of the cell.
5. The method of claim 1 , wherein the at least two molecular targets are selected from the group consisting of nucleic acid molecules, DNAs, RNAs, mRNAs, proteins, lipids, carbohydrates, and combinations thereof.
6. The method of claim 1 , further comprising:
removing a preceding plurality of probes from the cell in a second or subsequent hybridization.
7. The method of claim 6 , further comprising:
applying a plurality of probes to the cell after the preceding plurality of probes is removed from the cell.
8. The method of claim 1 , wherein the identifying step comprises:
locating, in the visual data for each second or subsequent hybridization, a visual signal within the defined region; and
identifying, in the visual data for a preceding hybridization, a second visual signal within the defined region, corresponding to the same molecular target.
9. The method of claim 1 , wherein the correlation function between the visual data for a second or subsequent hybridization and the visual data for a preceding hybridization is determined in a spatial domain.
10. The method of claim 1 , wherein the visual data comprise:
an image A of the defined region for the hybridization with pixel values; and
an image B of the defined region for the second hybridization with predetermined pixel values.
11. The method of claim 1 , wherein the correlation function between image A and image B is:
C
(
k
,
l
)
=
∑
m
=
0
M
-
1
∑
n
=
0
N
-
1
A
(
m
,
n
)
B
¯
(
m
-
k
,
n
-
l
)
,
(
i
)
where M and N are dimensions of images in x axis and y axis, respectively, A and B are the images, C is the correlation matrix with k and l as the spatial lag variables;
(ii) C(i,j)=F −1 {(m,n))F*(B(m,n))}, where F is the Fourier transform operation, A and B are the images, and C is the correlation matrix with i and j as the spatial lag variables; or
G
(
i
,
j
)
=
F
-
1
{
F
(
A
(
m
,
n
)
)
F
*
(
B
(
m
,
n
)
)
}
〈
A
(
m
,
n
)
〉
〈
B
(
m
,
n
)
〉
-
1
,
(
iii
)
where F is the Fourier transform operation, A and B are the images, and G is the correlation matrix with i and j as the spatial lag variables, which has been normalized.
12. The method of claim 11 , wherein the abundance level is
〈
N
1
2
〉
=
G
1
2
(
0
,
0
)
G
1
1
(
0
,
0
)
G
2
2
(
0
,
0
)
,
where G 12 (0,0) is the amplitude of the cross-correlation of image A and image B, G 11 (0,0) is the autocorrelation amplitude of image A, and G 22 (0,0) is the autocorrelation amplitude of image B.
13. The method of claim 1 , wherein the plurality of hybridizations {1, 2, . . . N} further comprises:
an additional hybridization during which an additional plurality of probes is applied to the cell, wherein the additional plurality of probes comprises at least two probes interacting with the same copy of the molecular target to provide a visual signal constituting the visual data for the additional hybridization, and wherein the visual signal differs from at least two preceding visual signals.
14. The method of claim 13 , further comprising:
applying the additional plurality of probes to the cell after two preceding pluralities of probes are removed from the cell.
15. The method of claim 2 , further comprising:
identifying a central region of each image, wherein the central region includes a correlation amplitude; and
applying background subtraction by summing all pixel values of each image, except those pixels within the central region.
16. The method of claim 2 , further comprising:
applying to each image, an image filter selected from the group consisting of deconvolution, deblurring, N-D filtering of multidimensional images, 2-D Gaussian filtering of images, 3-D Gaussian filtering of 3-D images, creating predefined 2-D filter, guided filtering of images, normalized 2-D cross-correlation, 2-D adaptive noise-removal filtering, 2-D median filtering, 2-D order-statistic filtering, local standard deviation of image, local range of image, local entropy of grayscale image, general sliding-neighborhood operations, extracting objects from binary image by size, extracting objects from binary image using properties, Pad array, 2-D frequency response, 2-D FIR filter using frequency sampling, 2-D FIR filter using frequency transformation, 2-D FIR filter using 1-D window method, 2-D FIR filter using 2-D window method, 2-D convolution matrix and combinations thereof.
17. The method of claim 1 , wherein the defined region has a density of the molecular targets of at least 0.5 to 10 molecules per μm 2 .
18. The method of claim 1 , further comprising:
assigning a color scheme to each abundance level of the molecular targets for each defined region in the plurality of defined regions and creating a color representation of distribution of the molecular targets within the cell.