Illumina Corporation

Emrah Kostem

Dietrich A Dehlinger

Ali Agah

Tracy Helen Fung

Frank J Steemers

Alex Aravanis

Sasan Amini

Byoungsok Jung

Alex So

Craig Lee Hetherington

Zhihong Zhang

Sante Gnerre

Xuyu Cai

Craig Hetherington

Helen Tracy Fung

In one embodiment, a sample surface of a biosensor includes pixel areas and holds a plurality of clusters during a sequence of sampling events such that the clusters are distributed unevenly over the pixel areas. In another embodiment, a biosensor has a sample surface that includes pixel areas and an array of wells overlying the pixel areas, the biosensor including two wells and two clusters per pixel area. The two wells per pixel area include a dominant well and a subordinate well. The dominant well has a larger cross section over the pixel area than the subordinate well. In yet another embodiment, an illumination system is coupled to a biosensor that illuminates the pixel areas with different angles of illumination during a sequence of sampling events, including, for a sampling event, illuminating each of the wells with off-axis illumination to produce asymmetrically illuminated well regions in each of the wells.

Systems and devices for high-throughput sequencing with semiconductor-based detection

We propose a neural network-implemented method for base calling analytes. The method includes accessing a sequence of per-cycle image patches for a series of sequencing cycles, where pixels in the image patches contain intensity data for associated analytes, and applying three-dimensional (3D) convolutions on the image patches on a sliding convolution window basis such that, in a convolution window, a 3D convolution filter convolves over a plurality of the image patches and produces at least one output feature. The method further includes beginning with output features produced by the 3D convolutions as starting input, applying further convolutions and producing final output features and processing the final output features through an output layer and producing base calls for one or more of the associated analytes to be base called at each of the sequencing cycles.

Base calling using convolution

The technology disclosed relates to determining tag signals from measured intensities for purposes of base calling in next-generation sequencing. In particular, the measured intensities are collected by light sensors in a sensor array directed to a sample surface including pixel areas and holding a plurality of clusters during a sequence of sampling events. Each light sensor is directed to and measuring intensity from one of the pixel areas during each sampling event. The method includes adjusting the measured intensities from a pixel in the pixel areas for background intensity based on variations in background levels of the light sensors in the sensor array and determining an intensity of a tag signal originating from the pixel based on the adjusted measured intensities of the pixel.

Systems and devices for signal corrections in pixel-based sequencing

A device for base calling is provided. The device includes a receptacle configured to hold a biosensor having a sample surface holding a plurality of clusters during a sequence of sampling events, an array of sensors sensing information from clusters disposed in corresponding pixel areas of the sample surface during the sampling events and generate sequences of pixel signals and a communication port configured to output the sequences of pixel signals. The device also includes a signal processor coupled to the communication port and configured to receive and process at least one pixel signal in the sequences of pixel signals that mixes light gathered from at least two clusters in a corresponding pixel area, and to base call each of the at least two clusters using the at least one pixel signal.

Semiconductor-based biosensors for base calling

The disclosed embodiments concern methods, apparatus, systems and computer program products for determining sequences of interest using unique molecular index (UMI) sequences that are uniquely associable with individual polynucleotide fragments, including sequences with low allele frequencies and long sequence length. In some implementations, the UMIs include both physical UMIs and virtual UMIs. In some implementations, the unique molecular index sequences include non-random sequences. System, apparatus, and computer program products are also provided for determining a sequence of interest implementing the methods disclosed.

Error suppression in sequenced DNA fragments using redundant reads with unique molecular indices (UMIS)

Base calling using three-dimentional (3D) convolution

We propose a neural network-based base caller that detects and accounts for stationary, kinetic, and mechanistic properties of the sequencing process, mapping what is observed at each sequence cycle in the assay data to the underlying sequence of nucleotides. The neural network-based base caller combines the tasks of feature engineering, dimension reduction, discretization, and kinetic modelling into a single end-to-end learning framework. In particular, the neural network-based base caller uses a combination ofD convolutions,D convolutions, and pointwise convolutions to detect and account for assay biases such as phasing and prephasing effect, spatial crosstalk, emission overlap, and fading.

Base calling using convolutions

Methods of determining a haplotype or partial haplotype of a DNA sample containing high molecular weight segments of genomic DNA are disclosed. Such methods may include sequencing DNA in an enriched DNA sample to produce a plurality of sequence reads, where some of the sequence reads contain a first allele of the first haplotype and other of the sequence reads contain a second allele of the first haplotype. Some methods align the sequence reads to a reference genome to produce aligned reads, where aligned reads from the first high molecular weight segment tend to cluster into islands on the reference genome. Some methods further determine distances separating adjacent aligned reads on the reference genome and select a first group of the aligned reads having separation distances to adjacent aligned reads that are smaller than a cutoff value. Using alleles from the first group of aligned reads, the method may define a first haplotype or first partial haplotype.

Haplotype resolved genome sequencing

In one embodiment, a method of determining tag signals from measured intensities, the measured intensities collected by light sensors in a sensor array directed to a sample surface, the sample surface including pixel areas and holding a plurality of clusters during a sequence of sampling events, each light sensor directed to and measuring intensity from one of the pixel areas during each sampling period includes adjustments for background intensity and crosstalk and taking into account signal decay and phasing/pre-phasing. Coefficients for the adjustments can be determined by gradient descent, using as ground truth base calling by the system being characterized or by using reliable base calling of well-characterized sample run through the system being characterized.

Systems and devices for characterization and performance analysis of pixel-based sequencing

A biosensor for base calling is provided. The biosensor comprises a sampling device, which includes a sample surface that has an array of pixel areas and a solid-state imager that has an array of sensors. Each sensor generates pixel signals in each base calling cycle. Each pixel signal represents light gathered in one base calling cycle from one or more clusters in a corresponding pixel area of the sample surface. The biosensor further comprises a signal processor configured for connection to the sampling device. The signal processor receives and processes the pixel signals from the sensors for base calling in a base calling cycle, and uses the pixel signals from fewer sensors than a number of clusters base called in the base calling cycle. The pixel signals from the fewer sensors include at least one pixel signal representing light gathered from at least two clusters in the corresponding pixel area.

Growing community of inventors

Menlo Park, CA, United States of America

Emrah Kostem