Deep learning pan-specific model for interpretable MHC-I peptide binding prediction with improved attention mechanism.

Accurate prediction of peptide binding affinity to the major histocompatibility complex (MHC) proteins has the potential to design better therapeutic vaccines. Previous work has shown that pan-specific prediction algorithms can achieve better prediction performance than other approaches. However, most of the top algorithms are neural networks based black box models. Here, we propose DeepAttentionPan, an improved pan-specific model, based on convolutional neural networks and attention mechanisms for more flexible, stable and interpretable MHC-I binding prediction. With the attention mechanism, our ensemble model consisting of 20 trained networks achieves high and more stabilized prediction performance. Extensive tests on IEDB's weekly benchmark dataset show that our method achieves state-of-the-art prediction performance on 21 test allele datasets. Analysis of the peptide positional attention weights learned by our model demonstrates its capability to capture critical binding positions of the peptides, which leads to mechanistic understanding of MHC-peptide binding with high alignment with experimentally verified results. Furthermore, we show that with transfer learning, our pan model can be fine-tuned for alleles with few samples to achieve additional performance improvement. DeepAttentionPan is freely available as an open-source software at https://github.com/jjin49/DeepAttentionPan.

Extending the leaf economics spectrum to decomposition: evidence from a tropical forest.

I investigated the relationship between leaf physiological traits and decomposition of leaf litter for 35 plant species of contrasting growth forms from a lowland tropical forest in Panama to determine whether leaf traits could be used to predict decomposition. Decomposition rate (k) was correlated with specific leaf area (SLA), leaf nitrogen (N), phosphorus (P), and potassium (K) across all species. Photosynthetic rate per unit mass (Amass) was not correlated with k, but structural equation modeling showed support for a causal model with significant indirect effects of Amass on k through SLA, N, and P, but not K. The results indicate that the decomposability of leaf tissue in this tropical forest is related to a global spectrum of leaf economics that varies from thin, easily decomposable leaves with high nutrient concentrations and high photosynthetic rates to thick, relatively recalcitrant leaves with greater physical toughness and defenses and low photosynthetic rates. If this pattern is robust across biomes, then selection for suites of traits that maximize photosynthetic carbon gain over the lifetime of the leaf may be used to predict the effects of plant species on leaf litter decomposition, thus placing the ecosystem process of decomposition in an evolutionary context.