Autobuild RNA (nautilus)

The “Autobuild RNA” task is used to either build an atomic model into an electron density map, or to rebuild a molecular replacement model to better fit the observations. The minimum required inputs are a the protein sequence, a set of observations, a set of Free-R flags, and either a set of phases or a molecular replacement model.

Input

image1

First the experimental data must be selected. A set of observations, a set of phases and a set of Free-R flags are required(1).

Next the contents of the asymmetric unit must be specified in the form of one or more chain sequences. Each sequence need only be provided once, however many times it appears in the asymmetric unit(2).

A partially built model may optionally be provided as a starting point for building(3).

Non-nucleotide features from the initial input model are preserved automatically. Nucleotide features will be preserved unless a more compelling trace is obtained for the same density.

Options

image2

The first group of options controls the model building pipeline. The only one of these you might normally change is the first: the number of build/refine iterations to be performed. If an initial model building job gives an incomplete model, rerunning with more cycles may help. While there is very rarely a benefit in using more than 20 cycles, we have seen rare examples where hundreds of cycles lead to a poor map being solved.

The final option - anisotropy correction - applies a correction to the data to remove the effect of the data being sharper in one direction than another. This should normally be on.

Advanced options

The “Resolution” option allows the use of higher resolution data, at a cost of run time.

Results

image3

The report summary describes how well the calculation has worked. Nautilus builds fragments of nucleotide chain, and tries to match those with the expected sequence. If the model is good, it will usually consist of a few long fragments which are grouped into as many chains as are present in the asymmetric unit. The chains will be largely complete, and all nucleotides built will be part of one of the chains. The refinement statistics are also an important indication of model accuracy.

Better models typically consist of more nucleotide units built, and few fragments (unless there is a lot of NCS). The results table provides a comprehensive analysis of the state of the final model.

Follow-on tasks include manual model building - use this if the model is nearly complete or if building has stalled.