Harnessing deep learning for population genetic inference

Huang, Xin; Rymbekova, Aigerim; Dolgova, Olga; Lao, Oscar; Kuhlwilm, Martin

doi:10.1038/s41576-023-00636-3

Cited by 15 publications

(9 citation statements)

References 157 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ML methods encompass a series of computer algorithm applications that refine their performance through experience ( Mitechell 1997 ). These approaches have gained popularity over the last few years for their efficacy in identifying population structure, making demographic inferences, and discerning signals of natural selection ( Huang et al 2023 ). Deep learning, a sophisticated class of machine-learning algorithms, has exhibited state-of-the-art performance across numerous applications involving large-scale population data ( Korfmann et al 2023 ).…”

Section: Overlooked Human Genetic Diversitymentioning

confidence: 99%

Biogeographic Perspectives on Human Genetic Diversification

Hünemeier

2024

Molecular Biology and Evolution

View full text Add to dashboard Cite

Modern humans originated in Africa 300,000 years ago, and before leaving their continent of origin, they underwent a process of intense diversification involving complex demographic dynamics. Upon exiting Africa, different populations emerged on the four other existing continents, shaped by the interplay of various evolutionary processes, such as migrations, founder effects, and natural selection. Within each region, continental populations, in turn, diversified and evolved almost independently for millennia. As a backdrop to this diversification, introgressions from archaic species contributed to establishing different patterns of genetic diversity in different geographic regions, reshaping our understanding of our species’ variability. With the increasing availability of genomic data, it has become possible to delineate the subcontinental human population structure precisely. However, the bias towards the genomic research focused on populations from the global North has limited our understanding of the real diversity of our species and the processes and events that guided different human groups throughout their evolutionary history. This perspective is part of a series of articles celebrating 40 years since our journal, Molecular Biology and Evolution, was founded (Russo et al. 2024). The perspective is accompanied by virtual issues, a selection of papers on human diversification published by Genome Biology and Evolution and Molecular Biology and Evolution.

show abstract

Section: Overlooked Human Genetic Diversitymentioning

confidence: 99%

Biogeographic Perspectives on Human Genetic Diversification

Hünemeier

2024

Molecular Biology and Evolution

View full text Add to dashboard Cite

show abstract

“…It proved valuable for analysing complex, high-dimensional genomics data and extracting previously unknown information. Examples of machine learning applications in the wider omics field range from the identification of DNA sequences (splice sites [32], promoters [33], enhancers [34]), nucleosome positioning [35], taxonomic annotation [36], microbial enterotyping [37], sequence errors learning [38], microbial host body site and subject classification [39], viral escape prediction [40], protein 3D structure estimation [41], evolutionary population genetics inference [42], and genomic selection [43].…”

Section: Machine Learning Solutions For Gwasmentioning

confidence: 99%

Machine Learning to Advance Human Genome-Wide Association Studies

Sigala,

Lagou,

Shmeliov

et al. 2023

Genes

View full text Add to dashboard Cite

Machine learning, including deep learning, reinforcement learning, and generative artificial intelligence are revolutionising every area of our lives when data are made available. With the help of these methods, we can decipher information from larger datasets while addressing the complex nature of biological systems in a more efficient way. Although machine learning methods have been introduced to human genetic epidemiological research as early as 2004, those were never used to their full capacity. In this review, we outline some of the main applications of machine learning to assigning human genetic loci to health outcomes. We summarise widely used methods and discuss their advantages and challenges. We also identify several tools, such as Combi, GenNet, and GMSTool, specifically designed to integrate these methods for hypothesis-free analysis of genetic variation data. We elaborate on the additional value and limitations of these tools from a geneticist’s perspective. Finally, we discuss the fast-moving field of foundation models and large multi-modal omics biobank initiatives.

show abstract

“…The advent of deep learning has largely facilitated the prediction of various layers of molecular phenotypes and the investigation of the regulatory information underlying genomic sequence [20][21][22][23] . Likewise, several deep learning models have been developed to predict ribosome profiling density along mRNA molecules using gene sequence as an input [24][25][26] .…”

Section: Introductionmentioning

confidence: 99%

Deep learning modeling of ribosome profiling reveals regulatory underpinnings of translatome and interprets disease variants

He,

Xiong,

Shi

et al. 2024

Preprint

View full text Add to dashboard Cite

Gene expression involves transcription and translation. Despite large datasets and increasingly powerful methods devoted to calculating genetic variants' effects on transcription, discrepancy between mRNA and protein levels hinders the systematic interpretation of the regulatory effects of disease-associated variants. Accurate models of the sequence determinants of translation are needed to close this gap and to interpret disease-associated variants that act on translation. Here, we present Translatomer, a multimodal transformer framework that predicts cell-type-specific translation from mRNA expression and gene sequence. We train Translatomer on 33 tissues and cell lines, and show that the inclusion of sequence substantially improves the prediction of ribosome profiling signal, indicating that Translatomer captures sequence-dependent translational regulatory information. Translatomer achieves accuracies of 0.72 to 0.80 for de novo prediction of cell-type-specific ribosome profiling. We develop an in silico mutagenesis tool to estimate mutational effects on translation and demonstrate that variants associated with translation regulation are evolutionarily constrained, both within the human population and across species. Notably, we identify cell-type-specific translational regulatory mechanisms independent of eQTLs for 3,041 non-coding and synonymous variants associated with complex diseases, including Alzheimer's disease, schizophrenia, and congenital heart disease. Translatomer accurately models the genetic underpinnings of translation, bridging the gap between mRNA and protein levels, and providing valuable mechanistic insights toward mapping "missing regulation" in disease genetics.

show abstract

Harnessing deep learning for population genetic inference

Cited by 15 publications

References 157 publications

Biogeographic Perspectives on Human Genetic Diversification

Biogeographic Perspectives on Human Genetic Diversification

Machine Learning to Advance Human Genome-Wide Association Studies

Deep learning modeling of ribosome profiling reveals regulatory underpinnings of translatome and interprets disease variants

Contact Info

Product

Resources

About