Generative protein language models are a natural way to design new proteins with desired functions.

A very promising direction in the field of black-box adversarial attacks are randomized search schemes for crafting adversarial examples [1, 23, 24]. Combining random search with specific update proposal distributions allows to achieve state-of-the-art black-box efficiency for different threat models such as` and `2 [1], `1 [25], `0, adversarial patches, and adversarial frames [24].

Anillustrativeexample inFigure 1(a)showcases thatthecorrelations analysis among amino acids (AAs) sites could reveal contacts or conservative regions in the folding structure.

Recently, another type of networks with stacked Transformer blocks emerged.

F s,iBs,i, n = 1,2,...,S, (10) where F s is the support features extracted by a pretrained ViT. Inspired by the multiple-object tracking within a single framework [21], in which different objects are represented by various identifications (i.e., learnable vectors) for simultaneously tracking, we add extra learnable tokens tothemeanfeatures formorediscriminativeprompts.

A.2 Documentation and intended uses We include a datasheet [1] in Section B. Detailed documentation on the precise structure and content OpenProteinSet is made available under the CC BY 4.0 license. The authors bear all responsibility in case of violation of rights. OpenProteinSet will continue to be hosted on RODA for the foreseeable future. A.7 Alignment tool settings For JackHMMer, we used -N 1 -E 0.0001 -incE 0.0001 -F1 0.0005 -F2 0.00005 -F3 0.0000005 and then capped outputs at depth 5000. B.1 Motivation For what purpose was the dataset created?

Each row of an MSA is a protein sequence.