[Power-Up] BCQEntanglemen: Genomic Error Correction and Resequencing with User Friendly QCNN

[ethraj2001]

Made By:

@erdabravest2001 @pranavkubc / @pranvkairon
@BestQuark @moreza14

Power Up Applications:

NVIDIA A100 GPU Timing

Challenges:

NVIDIA Challenge Quantum computing today! Amazon Braket Challenge Hybrid Quantum-Classical Computing Challenge Visualization Challenge

Summary:

“The goal of this project is to explore the use of quantum neural networks (QNNs) for correcting sequencing errors in genomics. Current sequencing technologies suffer from high error rates, which can cause significant problems in genomics research. Traditional error-correction methods rely on heuristic algorithms, which can be computationally expensive and often fail to identify all errors. Quantum computing, on the other hand, has the potential to provide significant speedup for many types of computational problems, including those in genomics.”

Background:

Genomics sequencing faces a significant challenge in error correction of readings, which can introduce errors and confound downstream analyses. While deep neural networks have enabled remarkable advancements in this field, errors present in sequence readings remain a major concern. Traditional algorithms, such as convolutional neural networks (CNNs), can be effective for error correction but may yield inconsistent results depending on the length of the sequence. Therefore, error correction remains an open problem in genomics sequencing that requires innovative solutions. Quantum convolutional neural networks (QCNNs) represent a promising avenue for addressing the challenge of error correction in genomics sequencing. QCNNs are uniquely positioned to provide novel solutions to this problem because they are capable of performing complex computations in parallel and can potentially leverage quantum entanglement to enhance feature detection and classification.

Objective:

This project aims to investigate the potential of quantum neural networks (QNNs) in genomic sequencing error correction. High error rates in current sequencing technologies can hinder genomics research, and traditional error-correction methods can be computationally expensive and limited in identifying errors [1-4]. QNNs can provide a significant speedup for computational problems in genomics. We will develop a QNN architecture specifically for error correction in genomic sequences and train it on a large dataset with known errors. We will evaluate the QNN's performance on a separate set of genomic sequences with known errors and compare it to traditional error-correction methods. We will also assess the accuracy and computational performance of the QNN-based method compared to other methods. Improvements in error correction can lead to more precise diagnosis and treatment of genetic diseases, better understanding of evolutionary processes, and more efficient agriculture and biotechnology. The development of QNN-based methods for genomic sequencing can open up broader applications of quantum computing in the life sciences.

Methods:

In this project, we will first develop a QNN architecture that is specifically tailored for error correction in genomic sequences. The QNN will be trained on a large dataset of genomic sequences with known errors. The training process will involve encoding the sequences as quantum states and using a quantum algorithm to optimize the parameters of the QNN. Once the QNN is trained, we will evaluate its performance on a separate set of genomic sequences with known errors. To assess the effectiveness of the QNN-based error correction, we will compare its performance to that of traditional error-correction methods, such as those based on Hidden Markov Models or statistical approaches and to NVIDIA Clara’s Genomic Reference Packages. We will evaluate the accuracy of the corrected sequences by comparing them to a reference genome. Additionally, we will examine the computational performance of the QNN-based method, including the time and resources required for training and correction. The potential impact of this project is significant, as improved error correction can lead to more accurate genomic sequencing, which can in turn enable more precise diagnosis and treatment of genetic diseases, better understanding of evolutionary processes, and more efficient agriculture and biotechnology. Furthermore, the development of QNN-based methods for genomic sequencing can pave the way for broader applications of quantum computing in the life sciences. In genomics sequencing, QCNNs can be trained to recognize patterns and identify errors in a given sequence of DNA. By encoding the sequence as a quantum state, QCNNs can potentially improve error correction accuracy and reduce false positives. Furthermore, QCNNs can be designed to be noise-resilient, making them more robust to errors and noise present in the sequencing process.One of the advantages of QCNNs over classical CNNs is their potential for speedup and parallelism, which can enable more efficient and accurate error correction for longer sequences. Additionally, QCNNs can be trained using quantum algorithms that are specifically designed for sequence analysis, which can further enhance their accuracy and efficiency. Overall, QCNNs represent a promising approach for addressing the challenge of error correction in genomics sequencing. By leveraging the unique properties of quantum computing, QCNNs have the potential to improve the accuracy, efficiency, and robustness of error correction, thereby facilitating more accurate downstream analyses and leading to new insights into the complex genetic code.

Results:

Any results of this project will be present in our project repo: https://github.com/erdabravest2001/BCQEntangleMen

References:

[ 1] "Canu: scalable and accurate long-read assembly via adaptive k-mer weighting and repeat separation" by Koren S, et al. Genome Research, 2017.

[2 ]"Deep learning for error correction in nanopore sequencing" by Liu et al. BMC Genomics, 2019.

[3] "DeepCov: predicting 3D genome folding using megabase-scale transfer learning from neural machine translation" by Hiranuma et al. Nature Communications, 2021.

[4] "Highly accurate read mapping with an extended Kalman filter" by Zhang et al. PLoS ONE, 2017.