TY - JOUR
T1 - Latent space of a small genetic network
T2 - Geometry of dynamics and information
AU - Seyboldt, Rabea
AU - Lavoie, Juliette
AU - Henry, Adrien
AU - Vanaret, Jules
AU - Petkova, Mariela D.
AU - Gregor, Thomas
AU - Francois, Paul
N1 - Funding Information:
ACKNOWLEDGMENTS. We thank Marianne Bauer, William Bialek, and Eric Wieschaus for discussion and comments. This work was supported, in part, by the US NSF, through the Center for the Physics of Biological Function (Grant PHY– 1734030); by NIH Grants R01GM097275, U01DA047730, and U01DK127429;
Funding Information:
by the Natural Sciences and Engineering Research Council of Canada, Discovery Grant program; and by the Simons Foundation, Mathematical Modeling of Living System program.
Publisher Copyright:
© 2022 National Academy of Sciences. All rights reserved.
PY - 2022/6/28
Y1 - 2022/6/28
N2 - The high-dimensional character of most biological systems presents genuine challenges for modeling and prediction. Here we propose a neural network-based approach for dimensionality reduction and analysis of biological gene expression data, using, as a case study, a well-known genetic network in the early Drosophila embryo, the gap gene patterning system.We build an autoencoder compressing the dynamics of spatial gap gene expression into a two-dimensional (2D) latent map.Theresulting 2Ddynamics suggests an almost linear model, with a small bare set of essential interactions.Maternally defined spatial modes control gap genes positioning, without the classically assumed intricate set of repressive gap gene interactions. This, surprisingly, predicts minimal changes of neighboring gap domains when knocking out gap genes, consistent with previous observations. Latent space geometries in maternal mutants are also consistent with the existence of such spatial modes. Finally, we show how positional information is well defined and interpretable as a polar angle in latent space. Our work illustrates how optimization of small neural networks onmedium-sized biological datasets is sufficiently informative to capture essential underlying mechanisms of network function.
AB - The high-dimensional character of most biological systems presents genuine challenges for modeling and prediction. Here we propose a neural network-based approach for dimensionality reduction and analysis of biological gene expression data, using, as a case study, a well-known genetic network in the early Drosophila embryo, the gap gene patterning system.We build an autoencoder compressing the dynamics of spatial gap gene expression into a two-dimensional (2D) latent map.Theresulting 2Ddynamics suggests an almost linear model, with a small bare set of essential interactions.Maternally defined spatial modes control gap genes positioning, without the classically assumed intricate set of repressive gap gene interactions. This, surprisingly, predicts minimal changes of neighboring gap domains when knocking out gap genes, consistent with previous observations. Latent space geometries in maternal mutants are also consistent with the existence of such spatial modes. Finally, we show how positional information is well defined and interpretable as a polar angle in latent space. Our work illustrates how optimization of small neural networks onmedium-sized biological datasets is sufficiently informative to capture essential underlying mechanisms of network function.
KW - Drosophila gap genes
KW - developmental biology
KW - dimensionality reduction
KW - machine learning
KW - systems biology
UR - http://www.scopus.com/inward/record.url?scp=85132682731&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85132682731&partnerID=8YFLogxK
U2 - 10.1073/pnas.2113651119
DO - 10.1073/pnas.2113651119
M3 - Article
C2 - 35737842
AN - SCOPUS:85132682731
SN - 0027-8424
VL - 119
JO - Proceedings of the National Academy of Sciences of the United States of America
JF - Proceedings of the National Academy of Sciences of the United States of America
IS - 26
M1 - e2113651119
ER -