Ammar has posted a preprint of his paper “rfPhen2Gen: A machine learning based association study of brain imaging phenotypes to genotypes” on arXiv. In this paper, Ammar evaluated random forest regression as a method to identify biologically relevant associations by learning models to predict SNPs from imaging features, using data for 518,484 SNPs and 56 brain imaging traits from the ADNI study.
Our paper “Restricted maximum-likelihood method for learning latent variance components in gene expression data with known and unknown confounders” has been published in G3. In this paper we analyse the mathematical structure of a class of statistical models for learning hidden factors influencing gene expression data and show that a new algorithm based on the analytical results is orders of magnitude faster than the standard algorithms for solving this class of models.
Congratulations to Ammar, whose paper “High-dimensional multi-trait GWAS by reverse prediction of genotypes” has been accepted for presentation at CIBB 2021!
Congratulations to Ramin, whose paper “A graph feature auto-encoder for the prediction of unobserved node features on biological networks” has been published in BMC Bioinformatics!
Congrats to Ramin for posting no less than two preprints on arXiv:
- Tesla-Rapture: A Lightweight Gesture Recognition System from mmWave Radar Point Clouds
- Integrating Sensing and Communication in Cellular Networks via NR Sidelink
Both papers result from a collaboration between Ramin and Dariush Salami at the Ambient Intelligence group at Aalto University, and introduce new graph neural network-based methods for analyzing point cloud data.
Great to see when students independently initiate projects and push them through to completion (exposing the limits of their supervisor’s knowledge in the process 🙂
Adriaan’s paper “Comparison between instrumental variable and mediation-based methods for reconstructing causal gene networks in yeast” has been posted on biorXiv. In this paper we used Findr to compare several methods for causal inference from genomics and transcriptomics data, using a recently published dataset of 1,012 segregants from a cross between two budding yeast strains.
Lingfei’s paper, “High-dimensional Bayesian network inference from systems genetics data using genetic node ordering” has been published in Frontiers in Genetics, in a Special Topic on Machine Learning and Network-Driven Integrative Genomics.
In this paper, we present a highly efficient approach for reconstructing Bayesian gene regulatory networks when prior information for the inclusion of edges exists or can be inferred from the available data. The method is implemented in the Findr software.
Pau’s paper “Model-based clustering of multi-tissue gene expression data” has been published in Bioinformatics. In this paper a method, called “revamp”, is introduced to find clusters (groups of genes with shared activity patterns) in multi-tissue data, where gene expression profiles are available from multiple tissues or organs sampled from the same group of individuals. Revamp improves existing methods by its ability to incorporate prior information on physiological tissue similarity, and by identifying a set of clusters, each consisting of a core set of genes conserved across tissues as well as differential sets of genes specific to one or more subsets of tissues. Revamp is implemented in the Lemon-Tree software.
Lingfei’s paper “Accurate wisdom of the crowd from unsupervised dimension reduction” has been published in Royal Society Open Science. In this paper it is shown that wisdom of the crowd, the collective intelligence derived from responses of multiple individuals to the same questions, is analogous to one-dimensional unsupervised dimension reduction in machine learning. This means that many of-the-shelf dimension reduction methods, such as good old PCA, can be repurposed as crowd-wisdom methods, usually with (much) better performance than existing default crowd-wisdom methods. Perhaps one of the more surprising results concerned the classification of skin images as being cancerous or not. As part of the hype surrounding deep learning, it was recently found that a deep neural network trained on 130,000 images was better at classifying a test set of 111 skin images than 21 individual dermatologists. However, we found that by doing a simple PCA of the predictions of these 21 dermatologists, they collectively outperformed the deep neural network. As The Economist put it in their recent ad, “not all intelligence is artificial”. In fact some of it is collective.