A generalizable framework to comprehensively predict epigenome, chromatin organization, and transcriptome

Zhang, Zhenhao; Qiu, Yiyang; Feng, Fan; Liu, Jie

doi:10.1101/2022.05.23.493129

Cited by 5 publications

(14 citation statements)

References 83 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We observe that the majority of model performance derives from the assay token, which leads to an increase in Pearson's r of approximately 50%. This is consistent with the aforementioned inter-assay distribution shifts and with prior work opting to use just an assay token [3] (see Section 4). Additionally, overall performance degrades with the inclusion of a tissue token, regardless of whether an assay token is used.…”

Section: Cgn Generalises Across Individuals Tissues Assays and Long-r...supporting

confidence: 87%

“…Although multiple recent models have focussed on epigenetic signals [25,4], the closest approach to ours is that of Epcot [3] (see Table 1), where multiple epigenetic assay types are predicted by cross-attending to assay embeddings. The authors also augment the fixed nature of the genome across cells with an accessibility signal (DNase-seq/ATAC-seq), which can be interpreted as a continuous cell type representation, while multitask pre-training followed by fine-tuning is used for optimisation.…”

Section: Discussionmentioning

confidence: 99%

“…To systematically evaluate the predictions made by CGN, we employ a more difficult holdout split than previous work [3,1,4]. We leverage the ENTEx dataset to assess generalisation across donors, epigenetic contexts, and sequence similarity (rather than simpler chromosomal validation).…”

Section: Cgn Generalises Across Individuals Tissues Assays and Long-r...mentioning

confidence: 99%

“…By mapping DNA sequences to targets computationally, epigenetic sequence models offer rapid mutagenesis across a range of experimental contexts [1,2]. When these models are constructed for interpolation, detailed epigenetic signals can be analysed in inaccessible tissues which are often missing or rare in existing databases [3,4]. Despite their recent success, DNA-based models for epigenetic feature prediction continue to add output tracks in a manner which limits scalability and bottlenecks performance.…”

Section: Introductionmentioning

confidence: 99%

“…Deep neural networks, comprising convolutional and Transformer-based modules [7,8], currently provide state of the art (SOTA) correlation when predicting epigenetic signals from the reference genome [3,2]. These models train across a large range of cell types and experimental contexts, where they recognise functionally-relevant regulatory motifs present in the DNA input sequence.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

AI in Respirology and Bronchoscopy

Deasy¹,

Colt²,

Kennedy³

2023

AI in Clinical Medicine

View full text Add to dashboard Cite

Sequence modelling approaches for epigenetic profile prediction have recently expanded in terms of sequence length, model size, and profile diversity. However, current models cannot infer on many experimentally feasible tissue and assay pairs due to poor usage of contextual information, limiting in silico understanding of regulatory genomics. We demonstrate that strong correlation can be achieved across a large range of experimental conditions by integrating tissue and assay embeddings into a Contextualised Genomic Network (CGN). In contrast to previous approaches, we enhance long-range sequence embeddings with contextual information in the input space, rather than expanding the output space. We exhibit the efficacy of our approach across a broad set of epigenetic profiles and provide the first insights into the effect of genetic variants on epigenetic sequence model training. Our general approach to context integration exceeds state of the art in multiple settings while employing a more rigorous validation procedure.

show abstract

Section: Cgn Generalises Across Individuals Tissues Assays and Long-r...supporting

confidence: 87%

Section: Discussionmentioning

confidence: 99%

Section: Cgn Generalises Across Individuals Tissues Assays and Long-r...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

AI in Respirology and Bronchoscopy

Deasy¹,

Colt²,

Kennedy³

2023

AI in Clinical Medicine

View full text Add to dashboard Cite

show abstract

Discriminative histone imputation using chromatin accessibility

Wen,

Zhong,

Zhang

et al. 2024

Preprint

View full text Add to dashboard Cite

Motivation: Histone modifications (HMs) play a crucial role in various biological processes, including transcription, replication, and DNA repair, and they have a significant impact on chromatin structure. However, annotating HMs across different cell types using experimental approaches alone is impractical due to cost and time constraints. Therefore, computational techniques are valuable for predicting HMs, complementing experimental methods, and enhancing our understanding of DNA characteristics and gene expression patterns. Results: In this study, we introduce a deep learning model called discriminative histone imputation using chromatin accessibility (dHICA). dHICA combines extensive DNA sequence information with chromatin accessibility data, utilizing the Transformer architecture. This allows dHICA to have a large receptive field, and more cell-type specific information, which can not only incorporate distal information across the entire genome, but also improve predictive accuracy and interpretability. Comparing performance with other models, we found that dHICA exhibits superior performance in both peak calling and signal predictions, especially in biologically salient regions such as gene elements. dHICA serves as a crucial tool for advancing our understanding of chromatin dynamics, offering enhanced predictive capabilities and interpretability, particularly in critical biological regions such as gene elements.

show abstract

The role of chromatin state in intron retention: a case study in leveraging large scale deep learning models

Daoud,

Ben-Hur

2024

Preprint

View full text Add to dashboard Cite

Complex deep learning models trained on very large datasets have become key enabling tools for current research in natural language processing and computer vision. By providing pre-trained models that can be fine-tuned for specific applications, they enable researchers to create accurate models with minimal effort and computational resources. Large scale genomics deep learning models come in two flavors: the first are large language models of DNA sequences trained in a self-supervised fashion, similar to the corresponding natural language models; the second are supervised learning models that leverage large scale genomics datasets from ENCODE and other sources. We argue that these models are the equivalent of foundation models in natural language processing in their utility, as they encode within them chromatin state in its different aspects, providing useful representations that allow quick deployment of accurate models of gene regulation. We demonstrate this premise by leveraging the recently created Sei model to develop simple, interpretable models of intron retention, and demonstrate their advantage over models based on the DNA langauage model DNABERT-2. Our work also demonstrates the impact of chromatin state on the regulation of intron retention. Using representations learned by Sei, our model is able to discover the involvement of transcription factors and chromatin marks in regulating intron retention, providing better accuracy than a recently published custom model developed for this purpose.

show abstract

A generalizable framework to comprehensively predict epigenome, chromatin organization, and transcriptome

Cited by 5 publications

References 83 publications

AI in Respirology and Bronchoscopy

AI in Respirology and Bronchoscopy

Discriminative histone imputation using chromatin accessibility

The role of chromatin state in intron retention: a case study in leveraging large scale deep learning models

Contact Info

Product

Resources

About