Identification of spurious and shadow ORFs

Gene prediction algorithms have limitations and can yield inaccurate ORFs predictions leading to spurious proteins, which can lead to spurious protein families. We decided to track the presence and the distribution of both “spurious” and “shadow” predicted ORFs in our clusters.

Methods

  • To detect eventual spurious ORFs, we screened our data set against the AntiFam database [1], which contains Pfam protein families “composed solely of spurious open reading frames (ORFs)”.
  • The shadow ORFs are artifacts produced during the identification of the coding region that defines an ORF. During this process, constraints are applied that can result in intervals and subintervals that overlap the coding region of the ORF, in one of the five possible reading frames (two on the same strand and three on the opposite strand) [2]. We identified the shadow ORFs in our dataset using the criteria applied in Yooseph et al., 2018 [3]. i) Two ORFs on the same strand are considered shadows if they overlap by at least 60 bps. ii) ORFs on opposite strands are identified as shadows if they overlap by at least 50 bps, and their ends of 3 ‘are within the intervals of the others, or if they overlap by at least 120 bps and the end of 5 ‘of one is in the interval of the other.

Results

Spurious ORFs

TOTAL: 53,324 (0.02%)

Distribution of spurious ORFs in the different data sets.

TARA Malaspina GOS OSD HMP
4,203 2,298 4,939 1,620 40,264

Shadows ORFs

TOTAL: 611,774 (0.2%)

Distribution of shadows ORFs in the different data sets.

TARA Malaspina GOS OSD HMP
157,688 40,762 66,245 70,632 276,447

Spurious and shadow ORFs in the clusters

We detected a total of 53,324 (0.02%) spurious ORFs distributed in 6,228 (0.02%) clusters.

Number of spurious ORFs in the clusters and in each project.

Spurious in clusters ≥ 10 members Spurious in clusters < 10 members > 1 Spurious in singletons
44,205 6,784 2,335

We identified 611,774 (0.2%) shadow ORFs distributed in 357,329 (1%) clusters.

Number of shadow ORFs in the clusters and in each project.

Shadows in clusters ≥ 10 members Shadows in clusters < 10 members > 1 Shadows in singletons
290,077 144,571 177,126



The scripts spur_shadow_orfs.sh and shadow_orfs.r identified the spurious and shadow ORFs in our dataset applying the criteria described above. The output is a tab-separated file containing the following fields:

  • orf_name
  • orf_length
  • cl_name
  • cl_size
  • prop_shadow_in_the_cluster
  • is.shadow
  • is.spurious

More info in the README.

[1] R. Y. Eberhardt, D. H. Haft, M. Punta, M. Martin, C. O’Donovan, and A. Bateman, “AntiFam: a tool to help identify spurious ORFs in protein annotation.,” Database: the journal of biological databases and curation, vol. 2012, p. bas003, Mar. 2012.

[2] S. Yooseph et al., “The Sorcerer II Global Ocean Sampling expedition: expanding the universe of protein families,” PLoS biology, vol. 5, no. 3, p. 16, 2007.

[3] S. Yooseph, W. Li, and G. Sutton, “Gene identification and protein classification in microbial metagenomic sequence data via incremental clustering.,” BMC bioinformatics, vol. 9, p. 182, Apr. 2008.

Let's Get In Touch!

[email protected]

functional-dark-side