Biology subjects

NaderiAlizadeh, N.

Publications and source records attributed to NaderiAlizadeh, N..

4 recordsLinked to original sources

bioRxiv2026

EvoPool: Evolution-Guided Pooling of Protein Language Model Embeddings

Protein language models (PLMs) encode amino acid sequences into residue-level embeddings that must be pooled into fixed-size representations for downstream protein-level prediction tasks. Although these embeddings implicitly reflect evolutionary constraints, existing pooling strategies operate on single sequences and do not explicitly leverage information from homologous sequences or multiple sequence alignments. We introduce EvoPool, a self-supervised pooling framework that integrates evolutionary information from homologs directly into aggregated PLM representations using optimal transport. Our method constructs a fixed-size evolutionary anchor from an arbitrary number of homologous sequences and uses sliced Wasserstein distances to derive query protein embeddings that are geometrically informed by homologous sequence embeddings. Experiments across multiple state-of-the-art PLM families on the ProteinGym benchmark show that EvoPool consistently outperforms standard pooling baselines for variant effect prediction, demonstrating that explicit evolutionary guidance substantially enhances the functional utility of PLM representations. Our implementation code is available at https://github.com/navid-naderi/EvoPool.

bioinformatics↗

bioRxiv2026

Context-Aware Protein Representations Using Protein Language Models and Optimal Transport

Proteins have different functions in different contexts. As a result, representations that take into account a proteins biological context would allow for a more accurate assessment of its functions and properties. Protein language models (PLMs) generate amino-acid-level (residue-level) embeddings of proteins and are a powerful approach for creating universal protein representations. However, PLMs on their own do not consider context and cannot generate context-specific protein representations. We introduce COPTER, a method that uses optimal transport to pool together a proteins PLM-generated residue-level embeddings using a separate context embedding to create context-aware protein representations. We conceptualize the residue-level embeddings as samples from a probabilistic distribution, and use sliced Wasserstein distances to map these samples against a context-specific reference set, yielding a contextualized protein-level embedding. We evaluate COPTERs performance on three downstream prediction tasks: therapeutic drug target prediction, genetic perturbation response prediction, and TCR-epitope binding prediction. Compared to state-of-the-art baselines, COPTER achieves substantially improved, near-perfect performance in predicting therapeutic targets across cell contexts. It also results in improved performance in predicting responses to genetic perturbations and binding between TCRs and epitopes. The implementation code is available at https://github.com/SahilP113/COPTER.

bioinformatics↗

bioRxiv2025

Protein Language Model-Aligned Spectra Embeddings for De Novo Peptide Sequencing

We consider the problem of de novo peptide sequencing in tandem mass spectrometry, where the goal is to predict the underlying peptide sequence given a spectrums fragment peaks and precursor information. We present PLMNovo, a constrained learning framework that leverages pre-trained protein language models (PLMs) to guide the training process. In particular, we cast peptide-spectrum matching as a constrained optimization problem that enforces alignment between spectrum and peptide embeddings produced by a spectrum encoder and a PLM, respectively. We use a Lagrangian primal-dual algorithm to train the spectrum encoder and the peptide decoder by solving the proposed constrained learning problem, while optionally fine-tuning the pre-trained PLM. Through numerical experiments on established benchmarks, we demonstrate that PLMNovo outperforms several state-of-the-art deep learning-based de novo sequencing algorithms.

bioinformatics↗

bioRxiv2024

Aggregating Residue-Level Protein Language Model Embeddings with Optimal Transport

Protein language models (PLMs) have emerged as powerful approaches for mapping protein sequences into embeddings suitable for various applications. As protein representation schemes, PLMs generate per-token (i.e., per-residue) representations, resulting in variable-sized outputs based on protein length. This variability poses a challenge for protein-level prediction tasks that require uniform-sized embeddings for consistent analysis across different proteins. Previous work has typically used average pooling to summarize token-level PLM outputs, but it is unclear whether this method effectively prioritizes the relevant information across token-level representations. We introduce a novel method utilizing optimal transport to convert variable-length PLM outputs into fixed-length representations. We conceptualize per-token PLM outputs as samples from a probabilistic distribution and employ sliced-Wasserstein distances to map these samples against a reference set, creating a Euclidean embedding in the output space. The resulting embedding is agnostic to the length of the input and represents the entire protein. We demonstrate the superiority of our method over average pooling for several downstream prediction tasks, particularly with constrained PLM sizes, enabling smaller-scale PLMs to match or exceed the performance of average-pooled larger-scale PLMs. Our aggregation scheme is especially effective for longer protein sequences by capturing essential information that might be lost through average pooling.

bioinformatics↗