Granted Patent B2

US 12,646,589 · App. 17/642,942 · Granted Jun 2, 2026

Inventors: Sofia Kyriazopoulou Panagiotopou (Redwood City, CA); Kai-How Farh (San Mateo, CA)

Assignee: Illumina, Inc.

G16B40/20G06F18/2137G06N3/063G16B30/00

Abstract

Claims (50)

1 . An artificial intelligence-based system, comprising:

one or more processors configures to execute processor-executable code; and

one or more memory structures configured to store processor-executable code, wherein the processor executable code, when executed, causes implementation of:

an input preparation module that accesses a sequence database and generates an input base sequence, wherein the input base sequence comprises

(i) a target base sequence with target bases, wherein the target base sequence is flanked by

(ii) a right base sequence with downstream context bases, and

(iii) a left base sequence with upstream context bases;

a sequence-to-sequence model that processes the input base sequence and generates an alternative representation of the input base sequence; and

an output module that processes the alternative representation of the input base sequence and produces at least one per-base output for each of the target bases in the target base sequence,

wherein the per-base output specifies, for a corresponding target base, signal levels of a plurality of epigenetic tracks.

2 . The artificial intelligence-based system of claim 1 , wherein the per-base output uses continuous values to specify the signal levels of the plurality of epigenetic tracks.

3 . The artificial intelligence-based system of claim 1 , wherein the output module further comprises a plurality of per-track processors, with each per-track processor corresponds to a respective epigenetic track in the plurality of epigenetic tracks and produces a signal level for a corresponding epigenetic track.

4 . The artificial intelligence-based system of claim 3 , wherein each per-track processor in the plurality of per-track processors processes the alternative representation of the input base sequence, generates a further alternative representation that is specific to a particular per-track processor, and produces the signal level for the corresponding epigenetic track based on the further alternative representation.

5 . The artificial intelligence-based system of claim 4 , wherein each per-track processor further comprises at least one residual block and a linear module and/or rectified linear unit (ReLU) module, and

wherein the residual block generates the further alternative representation and the linear module and/or the ReLU module processes the further alternative representation and produces the signal level for the corresponding epigenetic track.

6 . The artificial intelligence-based system of claim 3 , wherein each per-track processor in the plurality of per-track processors processes the alternative representation of the input base sequence and produces the signal level for the corresponding epigenetic track based on the alternative representation.

7 . The artificial intelligence-based system of claim 6 , wherein each per-track processor further comprises a linear module and/or a ReLU module.

8 . The artificial intelligence-based system of claim 1 , wherein the plurality of epigenetic tracks includes deoxyribonucleic acid (DNA) methylation changes (e.g., CpG), histone modifications, noncoding ribonucleic acid (ncRNA) expression, and chromatin structural changes (e.g., nucleosome positioning).

9 . The artificial intelligence-based system of claim 1 , wherein the plurality of epigenetic tracks includes deoxyribonuclease (DNase) tracks.

10 . The artificial intelligence-based system of claim 1 , wherein the plurality of epigenetic tracks includes histone 3 lysine 27 acetylation (H3K27ac) tracks.

11 . The artificial intelligence-based system of claim 1 , wherein the sequence-to-sequence model is a convolutional neural network.

12 . The artificial intelligence-based system of claim 11 , wherein the convolutional neural network further comprises groups of residual blocks.

13 . The artificial intelligence-based system of claim 12 , wherein each group of residual blocks is parameterized by a number of convolution filters in the residual blocks, a convolution window size of the residual blocks, and an atrous convolution rate of the residual blocks.

14 . The artificial intelligence-based system of claim 13 , wherein the size of convolution window varies between groups of residual blocks.

15 . The artificial intelligence-based system of claim 12 , wherein the convolutional neural network is parameterized by a number of residual blocks, a number of skip connections, and a number of residual connections.

16 . The artificial intelligence-based system of claim 12 , wherein each group of residual blocks produces an intermediate output by processing a preceding input, wherein dimensionality of the intermediate output is (I−[{(W−1)*D}*A])×N, where

I is dimensionality of the preceding input,

W is convolution window size of the residual blocks,

D is atrous convolution rate of the residual blocks,

A is a number of atrous convolution modules in the group, and

N is a number of convolution filters in the residual blocks.

17 . The artificial intelligence-based system of claim 16 , wherein the atrous convolution rate progresses non-exponentially from a lower residual block group to a higher residual block group.

18 . The artificial intelligence-based system of claim 17 , wherein atrous convolutions conserve partial convolution calculations for reuse as adjacent bases are processed.

19 . The artificial intelligence-based system of claim 12 , wherein each residual block further comprises at least one batch normalization module, at least ReLU module, at least one atrous convolution module, and at least one residual connection.

20 . The artificial intelligence-based system of claim 19 , wherein each residual block further comprises two batch normalization modules, two ReLU non-linearity modules, two atrous convolution modules, and one residual connection.

21 . The artificial intelligence-based system of claim 11 , wherein the convolutional neural network further comprises dimensionality altering convolution modules that reshape spatial and feature dimensions of a preceding input.

22 . The artificial intelligence-based system of claim 1 , wherein dimensionality of the input base sequence is (Cu+1+Cd)×4, where

Cu is a number of upstream context bases in the left base sequence,

Cd is a number of downstream context bases in the right base sequence, and

L is a number of target base in the target base sequence.

23 . The artificial intelligence-based system of claim 1 , wherein the sequence-to-sequence model is trained on training data that includes both coding and non-coding bases.

24 . An artificial intelligence-based system, comprising:

one or more processors configures to execute processor-executable code; and

one or more memory structures configured to store processor-executable code, wherein the processor executable code, when executed, causes implementation of:

a sequence-to-sequence model;

a plurality of per-track processors corresponding to a respective epigenetic track in a plurality of epigenetic tracks,

wherein each per-track processor further comprises at least one processing module (e.g., a residual block) and an output module (e.g., a linear module, a rectified linear unit (ReLU) module);

wherein the sequence-to-sequence model processes an input base sequence and generates an alternative representation of the input base sequence;

wherein each per-track processor's processing module processes the alternative representation and generates a further alternative representation that is specific to a particular per-track processor; and

wherein each per-track processor's output module processes the further alternative representation generated by its corresponding processing module and produces, as output, a signal level for a corresponding epigenetic track and for each base in the input base sequence.

Assignments (1)

Continuity (2)

References Cited (305)