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.
A new preprint, “High-dimensional Bayesian network inference from systems genetics data using genetic node ordering”, is available on bioRxiv. Bayesian networks are statistical models for gene regulatory networks, and their inference from large-scale omics data is a major problem in systems genetics. In this paper we present an algorithm to solve this problem that uses causal inference, topological sorting and variable selection, and that is much more efficient than traditional Markov chain Monte Carlo algorithms. The algorithm is implemented in the Findr and lassopv software packages.
We contributed two chapters to a Methods in Molecular Biology book on Gene Regulatory Networks: a chapter by Lingfei about the use of Findr for the inference of transcriptome-wide causal networks, and a chapter by Pau about the use of lemon-tree for the inference of differential module networks.
It’s been a long time in the making, but Sid’s work on using long read sequencing to determine expressed antigen diversity in Trypanosoma brucei infections has finally been posted on bioRxiv!
In this collaboration with Liam Morrison, we applied long read sequencing (PacBio) to VSG amplicons generated from blood extracted from mice infected with T. brucei. We found that long read sequencing is reliable for resolving allelic differences between VSGs, that there is significant expressed diversity (449 VSGs detected across 20 mice) and that there is a striking semi-reproducible pattern of expressed diversity across the timeframe of study.
Well done Sid and everyone else who contributed to this study!
There is a lot of research showing that genetic information in combination with gene expression data can be used to predict causal interactions between genes, on the basis that genetic variation among individuals causes gene expression variation but not vice versa (this PLOS CompBio article is a contribution to the field from our group and has links to earlier work). Anagha Joshi’s group asked if this principle could be extended to other contexts, and in a joint preprint “Causal gene regulatory network inference using enhancer activity as a causal anchor” an affirmative answer is given: variation of epigenetic activity at enhancer elements across multiple cell types or experimental treatments together with gene expression data also predicts causal interactions. The accompanying statistical methods have been implemented in our Findr software.
A preprint on “Model-based clustering of multi-tissue gene expression data” is available from arXiv. In this paper we present a Bayesian model-based clustering algorithm for large-scale, multi-tissue gene expression data, where expression profiles are obtained from multiple tissues or organs sampled from dozens to hundreds of individuals. Our model can incorporate prior information on physiological tissue similarity, and results in 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. The algorithm has been implemented in the Lemon-Tree software as a new “task”, revamp.
We have posted a preprint Wisdom of the crowd from unsupervised dimension reduction on arXiv. In this paper we show that one-dimensional unsupervised dimension reduction, such as principal component analysis and Isomap, can be used to derive consensus predictions from the responses of multiple individuals to the same questions, and performs better than existing solutions. This is relevant for crowd wisdom applications in the social and natural sciences, including data fusion, meta-analysis, crowd-sourcing, and committee decision making.
We have contributed two chapters to a forthcoming book on gene regulatory network inference edited by Vân Anh Huynh-Thu and Guido Sanguinetti.
Whole-transcriptome causal network inference with genomic and transcriptomic data (on bioRxiv) describes a protocol for reconstructing causal gene networks from genome-wide genotype and gene expression data using the Findr software.
Learning differential module networks across multiple experimental conditions (on arXiv) reviews the theory of module network inference and describes how differential module networks across multiple experimental conditions can be learned using the Lemon-Tree software.