esm

name	esm
description	ESM2 protein language model for embeddings and sequence scoring. Use this skill when: (1) Computing pseudo-log-likelihood (PLL) scores, (2) Getting protein embeddings for clustering, (3) Filtering designs by sequence plausibility, (4) Zero-shot variant effect prediction, (5) Analyzing sequence-function relationships. For structure prediction, use chai or boltz. For QC thresholds, use protein-qc.
license	MIT
category	design-tools
tags	["sequence-design","embeddings","scoring"]
source	https://github.com/adaptyvbio/protein-design-skills

ESM2 Protein Language Model

Prerequisites

Requirement	Minimum	Recommended
Python	3.8+	3.10
PyTorch	1.10+	2.0+
GPU VRAM	8GB	24GB (A10G)

How to run

Option 1: Modal

cd biomodals
modal run modal_esm2_predict_masked.py \
  --input-faa sequences.fasta \
  --out-dir embeddings/

Option 2: Python API

import torch, esm

model, alphabet = esm.pretrained.esm2_t33_650M_UR50D()
batch_converter = alphabet.get_batch_converter()
model = model.eval().cuda()

data = [("seq1", "MKTAYIAKQRQISFVK...")]
_, _, batch_tokens = batch_converter(data)

with torch.no_grad():
    results = model(batch_tokens.cuda(), repr_layers=[33])

embeddings = results["representations"][33]

ESM2 Models

Model	Parameters	Best For
esm2_t6_8M	8M	Fast screening
esm2_t33_650M	650M	Standard ✓
esm2_t36_3B	3B	Best quality

PLL Interpretation

Normalized PLL	Interpretation
> 0.2	Very natural sequence
0.0 - 0.2	Good, natural-like
-0.5 - 0.0	Acceptable
< -0.5	May be unnatural

Typical performance

Campaign	Time (A10G)	Cost (Modal)
100 sequences	5-10 min	~$1
1000 sequences	30-60 min	~$5

Throughput: ~100-200 sequences/minute with 650M model.

Next: Structure prediction with chai or boltz → protein-qc for filtering.