Machine Learning Methods for Exploring Sequence Determinants of 3D Genome Organization

Yang, Muyu; Ma, Jian

doi:10.1016/j.jmb.2022.167666

Cited by 19 publications

(8 citation statements)

References 73 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The most commonly used IML methods are post-hoc explanations, which are flexible and generally model-agnostic due to the fact that they are applied after the design and training of a prediction model. Feature importance methods are commonly used in computational biology applications (Novakovsky et al, 2022; Yang and Ma, 2022). These are methods that assign each input feature (e.g., a pixel in an cellular image or a DNA sequence feature) an importance value based on its contribution to the model prediction.…”

Section: Preliminaries On Iml Methodsmentioning

confidence: 99%

Best Practices for Interpretable Machine Learning in Computational Biology

Chen

Yang

Cui

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Advances in machine learning (ML) have enabled the development of next-generation prediction models for complex computational biology problems. These developments have spurred the use of interpretable machine learning (IML) to unveil fundamental biological insights through data-driven knowledge discovery. However, in general, standards and guidelines for IML usage in computational biology have not been well-characterized, representing a major gap toward fully realizing the potential of IML. Here, we introduce a workflow on the best practices for using IML methods to perform knowledge discovery which covers verification strategies that bridge data, prediction model, and explanation. We outline a workflow incorporating these verification strategies to increase an IML method's accountability, reliability, and generalizability. We contextualize our proposed workflow in a series of widely applicable computational biology problems. Together, we provide an extensive workflow with important principles for the appropriate use of IML in computational biology, paving the way for a better mechanistic understanding of ML models and advancing the ability to discover novel biological phenomena.

show abstract

Section: Preliminaries On Iml Methodsmentioning

confidence: 99%

Best Practices for Interpretable Machine Learning in Computational Biology

Chen

Yang

Cui

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Chromatin loop is a three-dimensional genome structure, by which two distant parts of genomic loci, comes in proximity, ignoring the presence of intervening DNA sequence. In chromatin, the formation of loop structure, controls the expression of speci c genes in transcription and replication level [2]. During transcription, CCCTC-binding factor (CTCF) protein can play an important role in loop formation.…”

Section: Introductionmentioning

confidence: 99%

Application of machine learning polymer models explaining hypokalemia in COVID-19 patients

BASU¹

2023

Preprint

View full text Add to dashboard Cite

Machine learning (ML) methods are used in genomics studies for elucidation of three-dimensional structure of chromatin in genome. In chromatin, the formation of loop in 3D structure, controls the expression of specific genes in transcription and replication level. There is a 100% sequence similarity between SARS CoV2 and human genome for seventy-one nucleotide sequence in chromosome 2. This short stretch of nucleotide sequence has been thoroughly analyzed with the help of UCSC Genome Brower. A few machine learning models for construction of 3D structure of human genome for this specific chromosomal region have been explored. In the specific chromosomal region, the presence of a protein coding gene (KCNJ3), enhancer, promoter and LTR region are identified. With the help of machine learning methods, the loop formation on the genome structure has been conformed. Topologically Associating Domains (TADs) in this specific chromosomal region are also predicted. The loop formation along the chromosomal region facilitates the binding of RNA polymerase II and CTCF in the promoter region of KCNJ3 gene. During transcription of KCNJ3 gene, epigenetic control of this gene expression occurs through acetylation in histone protein present in nucleosome complex. Thus, the rate of gene expression of KCNJ3 gene increases, which rises the K+ concentration inside the cells, resulting in hypokalemia in blood serum of COVID-19 patients.

show abstract

“…Deep learning models perform well in predicting enhancer activity [16,17], transcription factor binding [18], gene expression [19], and genome folding [20,21] from sequence, with newer models increasing scale and incorporating epigenetic assays to provide cell type-specific context [22][23][24]. We can probe these models as computational oracles to predict the behavior of DNA sequence at scales intractable experimentally [25]. Models have been applied to predict the impact of structural variants on human genome folding [20,26], confirm the importance of CTCF through computational mutagenesis [20], and resurrect the folding of Neanderthal genomes [27].…”

Section: Introductionmentioning

confidence: 99%

In silico discovery of repetitive elements as key sequence determinants of 3D genome folding

Gunsalus

Keiser

Pollard

2022

Preprint

View full text Add to dashboard Cite

Natural and experimental genetic variants can modify DNA loops and insulating boundaries to tune transcription, but it is unknown how sequence perturbations affect chromatin organization genome-wide. We developed an in silico deep-learning strategy to quantify the effect of any insertion, deletion, inversion, or substitution on chromatin contacts and systematically scored millions of synthetic variants. While most genetic manipulations have little impact, regions with CTCF motifs and active transcription are highly sensitive, as expected. However, our analysis also points to noncoding RNA genes and several families of repetitive elements as CTCF motif-free DNA sequences with particularly large effects on nearby chromatin interactions, sometimes exceeding the effects of CTCF sites and explaining interactions that lack CTCF. We anticipate that our available disruption tracks may be of broad interest and utility as a measure of 3D genome sensitivity and our computational strategies may serve as a template for biological inquiry with deep learning.

show abstract

Machine Learning Methods for Exploring Sequence Determinants of 3D Genome Organization

Cited by 19 publications

References 73 publications

Best Practices for Interpretable Machine Learning in Computational Biology

Best Practices for Interpretable Machine Learning in Computational Biology

Application of machine learning polymer models explaining hypokalemia in COVID-19 patients

In silico discovery of repetitive elements as key sequence determinants of 3D genome folding

Contact Info

Product

Resources

About