Fast and Sensitive Protein Alignment using DIAMOND
Nature Methods, 2014 · paper · Biomedicine
Introduces DIAMOND, an open-source algorithm that aligns sequencing reads against protein databases using double indexing and spaced seeds. The authors report it is 20,000 times faster than BLASTX on short reads with similar sensitivity.
Checked: PDF from uni-tuebingen.de, fetched 5 Oct 2026. Page numbers are those of that PDF.
1 finding
Reference or label Ref. 9 printed as "Burkhardt, S. & Kärkkäinen, J. Fundamenta Informaticae 23, 1001–1018 (2003)."
- p. 2 Reference 9
Fundamenta Informaticae 23, 1001–1018
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: that Burkhardt & Kärkkäinen (2003) appeared in Fundamenta Informaticae vol. 56, pp. 51–70. The p.2 reference list image shows volume 23, pages 1001–1018, year 2003, as quoted.
Second check: another model agreed
Report a problem with this finding
6 minor (typos and the like; most checked by the judge only)
Wording "The peak memory usage of DIAMOND-default was 100 GB" uses a configuration name not defined elsewhere; the paper otherwise names only DIAMOND-fast and DIAMOND-sensitive.
- p. 2 Main text, left column
The peak memory usage of DIAMOND-default was 100 GB
Judge's note: p.2 crop. Page 1 says four shapes are used 'by default', and the Methods command for DIAMOND-fast has no -sensitive flag, so the reader can infer that default means fast. The label is still inconsistent.
Second check: another model agreed
Report a problem with this finding
Wording The Methods list "RAPSearch2-default: rapsearch -z 48 -v 250 -e -1 -b 1", while the text and Fig. 1 call this configuration "RAPSearch2-sensitive".
- p. 4 Online Methods, Experimental study, parameter settings
RAPSearch2-default: rapsearch -z 48 -v
- p. 2 Main text and Figure 1 legend
RAPSearch2-sensitive
- p. 2 Main text, first paragraph
using a fast setting (−fast) and a sensitive setting (−sensitive)
- p. 4 Online Methods, Experimental study, parameter settings
RAPSearch2-default: rapsearch -z 48 -v 250 -e -1 -b 1
Judge's note: Checked the p.4 parameter list and the p.2 text and legend. No 'sensitive' RAPSearch2 configuration is defined; the paper never says the default run is the sensitive one. It can be inferred, since the fast run differs only by '-a T', but the labels are inconsistent.
Second check: another model agreed
Report a problem with this finding
Wording "about 100 GB ... This takes 100 s to generate in memory and about six times as long to read from disk at a typical read rate of 150 MB per second."
- p. 3 Online Methods, Double indexing, right column
This takes 100 s to generate in memory and about six times as long to read from disk at a typical read rate of 150 MB per second.
Judge's note: p.3 crop. 100 GB at 150 MB/s takes about 667 s, which is 6.7 times the 100 s build time. Since the index size is itself given as 'about 100 GB', 'about six times' is loose rather than clearly wrong. Noted as an observation.
Second check: another model agreed
Report a problem with this finding
Typo Fig. 1 e-value cutoff given as "≤0.001" in the legend but as "below 10−3" in the Online Methods (and introduction).
- p. 1 Introduction
with an expected value below 10−3
- p. 4 Online Methods, Experimental study
For Figure 1, we considered only alignments with a BLASTX e-value below 10−3
- p. 2 Figure 1 legend
Only alignments with an expected value of ≤0.001 are considered
Judge's note: Crops of the p.2 legend and the p.4 Methods confirm the non-strict and strict inequalities. Few, if any, alignments would have an e-value of exactly 0.001, so this is a minor consistency point only.
Second check: none (judge only)
Report a problem with this finding
Typo BLASTX command printed as "blastx _num_threads 48"; the option should be "-num_threads".
- p. 4 Online Methods, parameter settings
blastx _num_threads 48
- p. 4 Online Methods, parameter settings
blastx _num_threads 48 -evalue 0.1
- p. 4
BLASTX: blastx _num_threads 48 -evalue 0.1 -max_target_seqs
Judge's note: The p.4 image shows an underscore where every other option starts with a hyphen, so this is not an extraction artefact. Reviewer note 3 (-evalue 0.1 vs the 10^-3 threshold) is not a defect: the run used 0.1, and alignments were filtered at 10^-3 for Fig. 1 and 10^-5 for Supplementary Table 1 (p.4).
Second check: none (judge only)
Report a problem with this finding
Typo Tense slip: "We download a set of contigs from a microbial assembly" should read "We downloaded".
- p. 4 Online Methods, Experimental study, right column
We download a set of contigs
- p. 4 Online Methods, Experimental study
We download a set of contigs from a microbial assembly
Judge's note: The p.4 image confirms the wording; the surrounding sentences are in the past tense.
Second check: none (judge only)