By Brenda Gonzalez | 2018-07-19 11:59:30
In single-particle cryo-EM publications, in the methods sections you will typically encounter terms such as “box size”, “particle extraction”, and “auto-picking”. All of these terms are related to image processing with cryo-EM data. When micrographs are collected for a typical purified, single-particle sample, they contain your particle of interest in a wide range of views. In the image above borrowed from Mallick, et al GroEL is evenly dispersed in the field of view. The center panel shows what the raw particles look like when extracted from the micrograph. Essentially, this step is to cut out the particles of interest from the raw micrographs in a given box size specified in pixels. Generally, there is such a low signal/noise ratio in the raw micrographs that it makes it difficult to identify the particles. However, by collecting many hundreds of particles similar to what is seen in the center inset, it is then possible to average and categorize the particles into various views of the particle—this creates a higher contrast representation of each view as shown in the rightmost panel of the figure above. These images are referred to as 2D class averages.
However, a common mistake in this process usually happens in during particle extraction step. By making the box size too small, there is a risk of cutting out important signal. On the other hand, by using to big of a box size, wasting RAM becomes an issue—this can be an important aspect in image processing considering recent single-particle structures are known to require on the order of hundreds of thousands of particles.
So what is the appropriate box size for your data? Here are some key factors in choosing a box size:
- Pixel size matters for image processing calculations. As mentioned before by other cryo-EM software developers, pixel box size does significantly affect computational time. Below is a table indicating the optimal box sizes for image processing. The best sizes to use are even numbers with a low prime factors such as 2, 3, 5, and 7.
- CTF signal can be cuttoff in small box sizes. When doing CTF estimation, particles can recuperate high resolution data and improve the final 3D reconstruction. However, if your extracted particle box size is too small, it can cut out the CTF signal and CTF correction will not be done correctly—restricting your resolution. To avoid this issue, use a box size about 150 Å larger than the average diameter of your particle and choose a preferred box size from the table below.
- For example, if your target particle is 100 Å wide in the micrographs and your Å/pixel size is 0.8 Å, your preferred box size would be 260 Å, or 325 pixels. However, 325 pixels is not a preferred box size, so you could round down to 324 or up to 336.
Good box sizes: 16, 20, 24, 28, 32, 36, 40, 48, 56, 60, 64, 72, 80, 84, 96, 100, 108, 112, 120, 128, 140, 144, 160, 168, 180, 192, 196, 200, 216, 224, 240, 252, 256, 280, 288, 300, 320, 324, 336, 360, 384, 392, 400, 420, 432, 448, 480, 500, 504, 512, 540, 560, 576, 588, 600, 640, 648, 672, 700, 720, 756, 768, 784, 800, 840, 864, 896, 900, 960, 972, 980, 1000, 1008, 1024, 1080, 1120, 1152, 1176, 1200, 1260, 1280, 1296, 1344, 1372, 1400, 1440, 1500, 1512, 1536, 1568, 1600, 1620, 1680, 1728, 1764, 1792, 1800, 1920, 1944, 1960, 2000
Follow Us