[Bug]: NA values

[Yuan374]

Operating System

Unix (e.g., Ubuntu 20.04)

Version

No response

Python Version (optional)

3.7

Python Virtual Environment

Conda (or related)

Execution Environment

Local environment after installation of all external dependencies

Bug Description

What do NA values in some of the grids (the black ones) mean? How should I deal with them?

Steps to Reproduce

I was using the flags -tomap circular_variance -proj mollweide -s 10

Relevant Log Output

180	10	Inf	NA
180	20	NA	NA
180	30	NA	NA
180	40	NA	NA
180	50	NA	NA
180	60	NA	NA
180	70	0.606	NA
180	80	0.6	NA
180	90	0.608	NA
180	100	0.605	MET_6_A, GLU_7_A, LYS_8_A, SER_9_A
180	110	0.6	SER_3_A, SER_4_A, GLU_5_A, MET_6_A, GLU_7_A
180	120	0.562	MET_1_A, GLY_2_A, SER_3_A, SER_4_A, GLU_5_A
180	130	0.524	MET_1_A
180	140	0.41	NA
180	150	NA	NA
180	160	NA	NA
180	170	NA	NA
180	180	Inf	NA
190	10	Inf	NA
190	20	NA	NA
190	30	NA	NA
190	40	NA	NA
190	50	NA	NA
190	60	NA	NA
190	70	NA	NA
190	80	NA	NA

Additional context (optional)

No response

Confirmation

• I confirm I have searched for duplicates and reviewed the relevant documentation.

[nchenche]

Hello @Yuan374 ,

Thanks for using Surfmap. May I ask what is the size of your protein and does its shape has a peculiar form? Because NA values can typically happen for small proteins or not globular ones.

Also, what is the outcome for -s argument default value?

[Yuan374]

Hello @Yuan374 ,

Thanks for using Surfmap. May I ask what is the size of your protein and does its shape has a peculiar form? Because NA values can typically happen for small proteins or not globular ones.

Also, what is the outcome for -s argument default value?

Thanks for responding!

This particular example is ~40kD (uniprot Q84LG2), but I tried a few other examples and had similar issues. I tried the default -s 5 and other projection methods, but they all gave me grids with NA values. Does the program require a certain size or shape? Is there a way to get around having NA values?

[nchenche]

Dear @Yuan374 ,

Thank you for the information. I took a look at your structure on uniprot: it is mainly a long central helix with disordered segments scattered around. Unfortunately, this is typically the kind of case surfmap will not deal correctly with since it sounds far away from being globular.

Is there a way to get around having NA values?

Have you tried to increase the default radius of atoms of the protein (-rad argument defaults to 3.0A)? This will artificially increase the surface of your protein and thus should limit (but likely not eradicate) the number of NA values^*.

---

^* An increase of the surface of the proteins results in more particles generated around its surface. Each particles is assigned a value (computed or from b-factor) from the closest residue around it. Particles and their feature value are then projected on a 2D map from their spherical coordinates. This coarse 2D map (that can have holes), is then smoothed by dividing the map into cells; each cell is then averaged from the values of its neighbor cells (like average pooling in CNN). If a cell has no neighbor, then it will result in a NA value.