1. |
- Dwivedi, Sanjiv, et al.
(author)
-
Deriving disease modules from the compressed transcriptional space embedded in a deep autoencoder
- 2020
-
In: Nature Communications. - : NATURE PUBLISHING GROUP. - 2041-1723. ; 11:1
-
Journal article (peer-reviewed)abstract
- Disease modules in molecular interaction maps have been useful for characterizing diseases. Yet biological networks, that commonly define such modules are incomplete and biased toward some well-studied disease genes. Here we ask whether disease-relevant modules of genes can be discovered without prior knowledge of a biological network, instead training a deep autoencoder from large transcriptional data. We hypothesize that modules could be discovered within the autoencoder representations. We find a statistically significant enrichment of genome-wide association studies (GWAS) relevant genes in the last layer, and to a successively lesser degree in the middle and first layers respectively. In contrast, we find an opposite gradient where a modular protein-protein interaction signal is strongest in the first layer, but then vanishing smoothly deeper in the network. We conclude that a data-driven discovery approach is sufficient to discover groups of disease-related genes. The study of disease modules facilitates insight into complex diseases, but their identification relies on knowledge of molecular networks. Here, the authors show that disease modules and genes can also be discovered in deep autoencoder representations of large human gene expression datasets.
|
|
2. |
- Magnusson, Rasmus, 1992-, et al.
(author)
-
Deep neural network prediction of genome-wide transcriptome signatures – beyond the Black-box
- 2022
-
In: npj Systems Biology and Applications. - : Springer Nature. - 2056-7189. ; 8:1
-
Journal article (peer-reviewed)abstract
- Prediction algorithms for protein or gene structures, including transcription factor binding from sequence information, have been transformative in understanding gene regulation. Here we ask whether human transcriptomic profiles can be predicted solely from the expression of transcription factors (TFs). We find that the expression of 1600 TFs can explain >95% of the variance in 25,000 genes. Using the light-up technique to inspect the trained NN, we find an over-representation of known TF-gene regulations. Furthermore, the learned prediction network has a hierarchical organization. A smaller set of around 125 core TFs could explain close to 80% of the variance. Interestingly, reducing the number of TFs below 500 induces a rapid decline in prediction performance. Next, we evaluated the prediction model using transcriptional data from 22 human diseases. The TFs were sufficient to predict the dysregulation of the target genes (rho = 0.61, P < 10−216). By inspecting the model, key causative TFs could be extracted for subsequent validation using disease-associated genetic variants. We demonstrate a methodology for constructing an interpretable neural network predictor, where analyses of the predictors identified key TFs that were inducing transcriptional changes during disease.
|
|
3. |
- Magnusson, Rasmus, et al.
(author)
-
RNA-sequencing and mass-spectrometry proteomic time-series analysis of T-cell differentiation identified multiple splice variants models that predicted validated protein biomarkers in inflammatory diseases
- 2022
-
In: Frontiers in Molecular Biosciences. - : Frontiers Media SA. - 2296-889X. ; 9
-
Journal article (peer-reviewed)abstract
- Profiling of mRNA expression is an important method to identify biomarkers but complicated by limited correlations between mRNA expression and protein abundance. We hypothesised that these correlations could be improved by mathematical models based on measuring splice variants and time delay in protein translation. We characterised time-series of primary human naive CD4(+) T cells during early T helper type 1 differentiation with RNA-sequencing and mass-spectrometry proteomics. We performed computational time-series analysis in this system and in two other key human and murine immune cell types. Linear mathematical mixed time delayed splice variant models were used to predict protein abundances, and the models were validated using out-of-sample predictions. Lastly, we re-analysed RNA-seq datasets to evaluate biomarker discovery in five T-cell associated diseases, further validating the findings for multiple sclerosis (MS) and asthma. The new models significantly out-performing models not including the usage of multiple splice variants and time delays, as shown in cross-validation tests. Our mathematical models provided more differentially expressed proteins between patients and controls in all five diseases. Moreover, analysis of these proteins in asthma and MS supported their relevance. One marker, sCD27, was validated in MS using two independent cohorts for evaluating response to treatment and disease prognosis. In summary, our splice variant and time delay models substantially improved the prediction of protein abundance from mRNA expression in three different immune cell types. The models provided valuable biomarker candidates, which were further validated in MS and asthma.
|
|