Biomolecular simulation has developed into a mature research tool describing processes at a resolution often inaccessible to experiment. Moreover, computer experiments allow us to study the effect of modifications to the boundary conditions or the biomolecules themselves, which may be relatively cumbersome to perform experimentally. Over the years we have gained considerable experience in describing the interactions between cytochrome P450 enzymes and its substrates and inhibitors, where their remarkable flexibility seems to be a recurring theme. We will discuss how molecular simulations help to understand metabolism of xenobiotics, to predict binding affinities, and to characterize structurally much less studied Cytochome P450s, such as those in plants.

P450Atlas (https://p450atlas.org/), the recently launched online tool for automated subfamily assignment of P450s, currently (2 April 2025) has 164 068 distinct protein sequences in the database classified into 7 663 families and 21 060 subfamilies. Classification of P450s into families and subfamilies rely on multiple sequence alignments and sequence identity. However, how strongly does this classification of P450s relate to protein structure and activity?

The 3DM systems built by Bio-Prodict (https://3dm.bio-prodict.com/) are structure-based information systems of protein super families which rely on structure-based multiple alignments of X-ray structures of proteins belonging to targeted superfamilies to identify a “core” for which 90 % of structures contain at least three consecutive amino acids with alpha carbons ideally within 2.5 angstroms of the average position. Residues belonging to the “core” are assigned 3DM numbers, which differ from the numbers in the original sequences, but allow that for a given position in the core information from different proteins can be linked. Subfamilies are created based on diverse representative structures selected by 3DM. Protein sequences obtained through BLAST searches are sorted into these subfamilies based on alignment of the “core”. The Biocatalysis Group at UFS had P450 3DM systems built and updated by Bio-Prodict in 2009, 2011, 2013, 2021, 2023 and 2024. The latest 2024 P450 3DM iteration has 175 334 aligned sequences sorted into 164 structure-based subfamilies with 387 “core” positions. Using P450Atlas, and with the assistance of Domique Gront and David Nelson, we assigned for the first time in 2025 full CYP names to all the structures representing the 164 structure-based subfamilies. This has allowed us to investigate the relationship between P450 classification and structure and how these relate to activity.

However, 3DM systems are currently limited by the availability of X-ray structures, which are for some P450 families over abundant while for others limited, because it is difficult, perhaps impossible, to obtain crystals. Thus 126 of the structure-based 3DM subfamilies are from Bacteria, while only 38 are from Eukaryotes. In comparison P450Atlas has assigned 4 665 CYP subfamilies to Bacteria and a total of 16 395 to Eukaryotes. The lack of X-ray structures from Eukaryotes might be addressed by including models created by AlphaFold3 (https://alphafoldserver.com/) into the 3DM P450 system. The accuracy of models created by AlphaFold3 has therefore been benchmarked against experimentally determined structures, and the feasibility of using AlphaFold3 models as templates for structure-based 3DM subfamilies evaluated, highlighting potential pitfalls.

P450Atlas has been recently published as a comprehensive and authoritative resource for cytochrome P450 (CYP) nomenclature, listing all known families and subfamilies along with their corresponding published sequences. In addition to its curated database, P450Atlas offers a robust sequence-based naming service that assigns CYP identifiers using HMM profiles and a refined alignment protocol. Although launched this February, the naming service received nearly 1,000 sequence submissions in April alone.
In this contribution, we present new developments that significantly expand the functionality of P450Atlas. Specifically, we have integrated structural data from the Protein Data Bank (PDB), incorporating information from 1,451 deposited P450 structures. These structures were also used to validate the performance of the naming service. Remarkably, the server correctly assigned CYP identifiers in all but six cases where only the subfamily was misassigned. In each of these, the input sequence showed borderline similarity to known subfamilies, with sequence identity near or below the 56% threshold. In contrast, only approximately 53% of the corresponding entries in the PDB or their associated publications had accurate CYP annotations, underscoring the need for a standardized approach.
To further assess the accuracy of the naming service, we analyzed all P450 entries in the SwissProt database. Representative P450 structures were used to refine the HMM profiles in order to improve domain detection and reduce the risk of false positives. As a result, we were able to accurately and reliably identify nearly two thousands CYP proteins in SwissProt.
We anticipate that these enhancements will further increase the utility of P450Atlas for the research community and support consistent, structure-aware annotation of cytochrome P450 sequences.

The proclamation that all the world’s DNA sequences will be represented soon by at least one genome per genus, will allow the complete annotation of all P450 families on earth. The problem is quite large, but it can be broken down to phylum level and lower taxonomic ranks in cases of very large phyla like arthropods. Humans and even vertebrates are a miniscule part of the whole covering about 10 million eukaryotes (mostly insects). There are approximately 82 eukaryotic phyla and about 45 prokaryotic phyla. The method to find all families in a phylum is to annotate whole genomes and keep track of new families as more genomes are added. At some point no new P450 families will be found. This is the point of saturation. Vertebrates have reached this point at 19 families. Land plants are saturated at 291 families. Angiosperms have only 57 families. Large taxa like insects are not yet saturated even though we have named 1066 P450 families from insects. Progress on this annotation will be presented for eukaryotes. Currently 11,406 P450 families are named.