Compute protein embeddings and PLL scores using the ESM2 language model.
npx skills add https://github.com/adaptyvbio/protein-design-skills --skill esm-adaptyvbio
Please help me install this Agent Skill. Skill: esm Source: https://github.com/adaptyvbio/protein-design-skills/tree/main/skills/esm Command: npx skills add https://github.com/adaptyvbio/protein-design-skills --skill esm-adaptyvbio
This Skill provides fast access to protein embeddings and sequence evaluation by applying the ESM2 language model, enabling data-driven analysis for design and evaluation.
Use Case: design a batch of sequences, generate embeddings, rank by PLL, and filter for plausibility before downstream analysis.
Use the ESM2 skill to generate embeddings and PLL for input sequences. Run the modal workflow: modal run modal_esm2_predict_masked.py --input-faa sequences.fasta --out-dir embeddings/ Or load esm2 via Python API to process batches in your own script.
High-intent search queries and answers about installing and using this skill.
Compute ESM2 embeddings for protein sequences by running the biomodals modal workflow via the modal_esm2_predict_masked.py script, passing your input FASTA file to generate vector representations for downstream clustering and similarity analysis.
Score protein sequence plausibility using pseudo-log-likelihood (PLL) by running the ESM2 language model workflow, which evaluates sequence naturalness to help rank and filter design batches before downstream analysis.
Zero-shot variant effect prediction with ESM2 leverages the pre-trained language model to evaluate mutation impacts across design datasets without requiring additional training, using pseudo-log-likelihood scores to assess sequence changes.
Running ESM2 sequence scoring requires Python 3.8 or higher and PyTorch, along with the biomodals modal workflow environment, to execute the modal_esm2_predict_masked.py script end-to-end.
Rank protein designs by sequence plausibility by computing pseudo-log-likelihood scores via the ESM2 language model, allowing you to filter designed sequences based on their naturalness before functional analysis.
ESM2 embeddings can be used for clustering protein sequences by generating vector representations through the modal workflow, enabling similarity analyses and downstream modeling for your design datasets.
adaptyvbioDesign protein sequences from structural backbones using ProteinMPNN inverse folding.
adaptyvbioRank binder designs using ipSAE scores from AlphaFold2, AlphaFold3, and Boltz1 predictions.
zongtingweiScore protein sequences and generate embeddings using the ESM2 model.
junior1pCompute ESM2 embeddings and pseudo-log-likelihood scores for protein sequences.
