Method for diagnosing cancer and predicting cancer type by using terminal sequence motif frequency and size of cell-free nucleic acid fragment
Disclosed is a method for diagnosing cancer and predicting a cancer type using fragment end motif frequencies and sizes of cell-free nucleic acid, and more preferably, to a method for diagnosing cancer and predicting a cancer type by extracting nucleic acids from a biological sample to obtain sequence information, acquiring fragment end motif frequencies and sizes of nucleic acids based on the aligned reads, converting the fragment end motif frequencies and sizes of nucleic acids into vectorized data, inputting the vectorized data to a trained artificial intelligence model and analyzing a resulting calculated value. The method includes generating vectorized data and analyzing the same using an AI algorithm and thus is useful due to high sensitivity and accuracy thereof even in the case of low read coverage.
1. A method for measuring fragment end motif frequency and size of nucleic acids, the method comprising:
a) extracting nucleic acids from a biological sample, the extracted nucleic acids comprising cell free DNA (cfDNA);
b) sequencing the nucleic acids to a depth of at least 1 million reads to obtain sequence reads;
c) aligning the sequence reads with a reference genome database, wherein the length of the reads is from 5 to 5,000 bp;
d) acquiring end motif frequencies and sizes of nucleic acid fragments based on the aligned sequence reads, wherein the end motif frequencies are selected from a total of 256 4-mer motifs, and wherein the nucleic acid fragments comprise fragment sizes ranging from 90 to 250 base pairs;
e) generating vectorized data using the end motif frequencies and sizes of nucleic acid fragments, wherein the vectorized data is expressed by a type of an end motif of the nucleic acid fragments and the size of the nucleic acid fragments plotted on a single or multidimensional axis;
f) inputting the generated vectorized data to a trained artificial intelligence model, wherein the artificial intelligence model transforms the vectorized data of step (e) into an output value, wherein the artificial intelligence model has been trained to adjust the output value to about 1 if there is cancer or to adjust the output value to about 0 if there is no cancer; and
g) determining whether or not cancer develops, wherein the determining comprises comparing the output value of step (f) with a cut-off value to determine whether or not cancer develops, wherein if the output value is greater than a cut-off value, it is determined that there is cancer, and if the output value is less than the cut-off value, it is determined that there is no cancer,
wherein the artificial intelligence model in step (f) has been trained to distinguish between vectorized data of a healthy subject and vectorized data of a cancer patient.
2. The method according to claim 1 , wherein step (a) comprises:
a-i. obtaining nucleic acids from the blood, semen, vaginal cells, hair, saliva, urine, oral cells, amniotic fluid containing placental cells or fetal cells, tissue cells, or a mixture thereof;
a-ii. removing proteins, fats, and other residues from the collected nucleic acids using a salting-out method, a column chromatography method, or a bead method to obtain purified nucleic acids;
a-iii. producing a single-end sequencing or paired-end sequencing library for the purified nucleic acids or nucleic acids randomly fragmented by an enzymatic digestion, pulverization, or Hydroshear method;
a-iv. reacting the produced library with a next-generation sequencer; and
a-v. obtaining sequence reads of the nucleic acids in the next-generation sequencer.
3. The method according to claim 1 , wherein the end motif of each nucleic acid fragment in step (d) has a sequence pattern of 2 to 30 bases at both ends of the nucleic acid fragment.
4. The method according to claim 1 , wherein the frequency of end motifs of the nucleic acid fragments in step (d) corresponds to the number of motifs detected in all the nucleic acid fragments.
5. The method according to claim 1 , wherein the size of each nucleic acid fragment in step (d) corresponds to the number of bases from the 5′ end to the 3′ end of the nucleic acid fragment.
6. The method according to claim 1 , wherein the vectorized data in step (d) is expressed by the type of the end motif of the nucleic acid fragment plotted on an X-axis and the size of the nucleic acid fragment plotted on a Y-axis.
7. The method according to claim 6 , wherein the vectorized data further comprises a sum of frequencies for end motifs of nucleic acid fragments and a sum of frequencies for sizes of nucleic acid fragments.
8. The method according to claim 1 , wherein the artificial intelligence model is a convolutional neural network (CNN).
9. The method according to claim 8 , wherein, a loss function for performing binary classification is represented by Equation 1 below and a loss function for performing multi-class classification is represented by Equation 2 below:
Equation
1
:
Binary
classification
loss
(
model
(
x
)
,
y
)
=
-
1
n
[
∑
i
=
1
n
(
y
i
log
(
model
(
x
i
)
)
+
(
1
-
y
y
)
log
(
1
-
model
(
x
t
)
)
)
]
Model (x i )=Artificial intelligence model output in response to i th input
y=Actual label value
n=Number of input data
Equation
2
:
Multi
-
class
classification
loss
(
model
(
x
)
,
y
)
=
-
1
n
∑
i
=
1
n
(
∑
j
=
1
c
(
y
ij
log
(
model
(
x
i
)
)
j
)
Model (x i ) j =j th artificial intelligence model output in response to i th input
y=Actual label value
n=Number of input data
c=Number of classes.
10. The method according to claim 1 , wherein the output value resulting from analysis of the input vectorized data by the artificial intelligence model in step (f) is a deep probability index (DPI).
11. The method according to claim 1 , wherein the cut-off value of step (e) is 0.5 and a determination is made that cancer has developed when the output value is 0.5 or more.
12. The method according to claim 1 , wherein the artificial intelligence model is a deep neural network (DNN).
13. The method according to claim 1 , wherein the artificial intelligence model is a recurrent neural network (RNN).
14. A non-transitory computer-readable storage medium comprising an instruction configured to be executed by a processor for diagnosing cancer through the following steps comprising:
(a) extracting nucleic acids from a biological sample to obtain sequence reads, the extracted nucleic acids comprising cell free DNA (cfDNA);
(b) aligning the sequence reads with a reference genome database;
(c) acquiring end motif frequencies and sizes of nucleic acid fragments based on the aligned sequence reads, wherein the end motif frequencies are selected from a total of 256 4-mer motifs, and wherein the nucleic acid fragments comprise fragment sizes ranging from 90 to 250 base pairs;
(d) generating vectorized data using the motif frequencies and sizes of nucleic acid fragments;
(e) inputting the generated vectorized data into a trained artificial intelligence model, wherein the artificial intelligence model transforms the vectorized data of step (d) into an output value, wherein the artificial intelligence model has been trained to adjust the output value to about 1 if there is cancer or to adjust the output value to about 0 if there is no cancer; and
(f) determining whether or not cancer develops, wherein the determining comprises comparing the output value of step (e) with a cut-off value to determine whether or not cancer develops, wherein if the output value is greater than a cut-off value, it is determined that there is cancer, and if the output value is less than the cut-off value, it is determined that there is no cancer,
wherein the artificial intelligence model in step (f) has been trained to distinguish between vectorized data of a healthy subject and vectorized data of a cancer patient.
15. The non-transitory computer-readable storage medium of claim 14 , further comprising predicting a cancer type, wherein the output value in step (e) comprises a deep probability index (DPI), wherein predicting a cancer type comprises comparing the DPI calculated for multiple cancer types, wherein the cancer type showing the highest DPI is predicted to be the cancer type of the sample.
16. A method for measuring fragment end motif frequency and size of nucleic acids, the method comprising:
a) extracting nucleic acids from a biological sample, the extracted nucleic acids comprising cell free DNA (cfDNA);
b) sequencing the nucleic acids to a depth of at least 1 million reads to obtain sequence reads;
c) aligning the sequence reads with a reference genome database, wherein the length of the reads is from 5 to 5,000 bp;
d) acquiring end motif frequencies and sizes of nucleic acid fragments based on the aligned sequence reads, wherein the end motif frequencies are selected from a total of 256 4-mer motifs, and wherein the nucleic acid fragments comprise fragment sizes ranging from 90 to 250 base pairs;
e) generating vectorized data using the end motif frequencies and sizes of nucleic acid fragments, wherein the vectorized data is expressed by a type of an end motif of the nucleic acid fragments and the size of the nucleic acid fragments plotted on a single or multidimensional axis;
f) inputting the generated vectorized data to a trained artificial intelligence model, wherein the artificial intelligence model transforms the vectorized data of step (e) into an output value, wherein the artificial intelligence model has been trained to adjust the output value to about 1 if there is cancer or to adjust the output value to about 0 if there is no cancer;
g) determining whether or not cancer develops, wherein the determining comprises comparing the output value of step (f) with a cut-off value to determine whether or not cancer develops, wherein if the output value is greater than a cut-off value, it is determined that there is cancer, and if the output value is less than the cut-off value, it is determined that there is no cancer; and
h) predicting a cancer type, wherein the output value in step (f) comprises a deep probability index (DPI), wherein predicting a cancer type comprises comparing the DPI calculated for multiple cancer types, wherein the cancer type showing the highest DPI is predicted to be the cancer type of the sample,
wherein the artificial intelligence model in step (f) has been trained to distinguish between vectorized data of a healthy subject and vectorized data of a cancer patient.