Abstract
Determination of functions for poorly characterized genes is crucial for understanding biological processes and studying human diseases. Functionally associated genes are often gained and lost together through evolution. Therefore identifying co-evolution of genes can predict functional gene-gene associations. We describe here the full statistical model and computational strategies underlying the original algorithm, CLustering by Inferred Models of Evolution (CLIME 1.0) recently reported by us [Li et al., 2014]. CLIME 1.0 employs a mixture of tree-structured hidden Markov models for gene evolution process, and a Bayesian model-based clustering algorithm to detect gene modules with shared evolutionary histories (termed evolutionary conserved modules, or ECMs). A Dirichlet process prior was adopted for estimating the number of gene clusters and a Gibbs sampler was developed for posterior sampling. We further developed an extended version, CLIME 1.1, to incorporate the uncertainty on the evolutionary tree structure. By simulation studies and benchmarks on real data sets, we show that CLIME 1.0 and CLIME 1.1 outperform traditional methods that use simple metrics (e.g., the Hamming distance or Pearson correlation) to measure co-evolution between pairs of genes.
Introduction
In biological research, the "guilt by association" principle is a cornerstone for functional annotation. When a gene's function is unknown, researchers often look at its "friends" — genes with which it frequently interacts or co-occurs. Phylogenetic profiling is a powerful method for identifying these associations by tracking the presence (1) or absence (0) of genes across a diverse set of species. The underlying assumption is that genes involved in the same biological pathway will be gained or lost together throughout evolutionary history.
Traditional approaches to phylogenetic profiling rely on simple distance-based metrics, such as Hamming distance or Pearson correlation, to quantify the similarity between gene profiles. However, these methods often fail to account for the hierarchical, tree-like nature of evolution, treating each species as an independent observation. This can lead to significant biases and sensitivity to noise in genomic data.
The CLIME Framework
To address these limitations, we developed CLIME (CLustering by Inferred Models of Evolution). CLIME utilizes a Bayesian mixture of tree-structured Hidden Markov Models (HMMs) to group genes into Evolutionary Conserved Modules (ECMs). Unlike linear HMMs, the tree-structured HMM follows the branches of a phylogenetic tree, where nodes represent species and transitions represent evolutionary gain or loss events.
Key components of the CLIME model include:
- Tree-structured HMM: Instead of a linear sequence, the HMM follows the branches of a phylogenetic tree. Each node represents a species (extant or ancestral), and the "hidden state" is the presence or absence of a gene.
- Dirichlet Process (DP) Prior: This non-parametric approach allows the model to automatically determine the number of gene clusters from the data, rather than requiring the user to specify a number upfront.
- Gibbs Sampling: A Markov Chain Monte Carlo (MCMC) technique used to sample the posterior distribution, enabling the estimation of cluster assignments and evolutionary parameters.
- Handling Uncertainty (CLIME 1.1): While CLIME 1.0 assumes a fixed phylogenetic tree, version 1.1 incorporates a prior distribution over multiple tree structures, making the model robust to uncertainties in evolutionary history.
Results and Validation
The performance of CLIME was evaluated using both simulated datasets and real-world biological benchmarks, including the human genome across 139 eukaryotic species. In simulations, CLIME 1.1 consistently achieved higher Adjusted Rand Index (ARI) scores compared to traditional metrics, particularly in scenarios with high data noise or tree misspecification.
Key findings include:
- Complex I Discovery: CLIME correctly partitioned the 44 genes of Mitochondrial Complex I into functional sub-clusters and identified essential assembly factors during the expansion phase.
- Ciliary Disease Validation: In an analysis of genes related to ciliary dyskinesia, CLIME identified 100 novel candidates, 50 of which were already validated in the "Ciliome" database, demonstrating high predictive power.
Discussion and Future Work
By modeling the tree structure, CLIME recognizes that a gene loss in a common ancestor is a single event, whereas traditional metrics would count it as multiple independent losses in all descendant species, leading to skewed results. Furthermore, by treating the phylogenetic tree as a distribution rather than a fixed fact, CLIME 1.1 avoids overfitting to potentially incorrect evolutionary histories.
Despite its success, the model currently assumes a fixed observation error rate and uniform noise across species. Future iterations could incorporate species-specific mis-observation rates and improved computational strategies to handle the intensive requirements of genome-wide Gibbs sampling.
Summary updated