ExUniProt-answers

The numbers are found using UniProt 2025_03 on Sep 16, 2025

Simple text mining

QUESTION 1.1:

1. How many hits do you find?
   35,870
2. How many of these hits are from Swiss-Prot?
   1,774
3. Can you identify the correct hit (i.e. see which one is actually human insulin and not something else)?
   It's P01308 / INS_HUMAN (not necessarily the top hit, but still on the first page).

QUESTION 1.2: How many hits are now left? How many of these are from Swiss-Prot?
2,375 and 1,157

QUESTION 1.3: How many hits are now left? How many of these are from Swiss-Prot?
217 and 60, search string: (organism_id:9606) AND (protein_name:insulin)

QUESTION 1.4: How many hits are now left? How many of these are from Swiss-Prot?
68 and 25, search string: (organism_id:9606) AND (protein_name:insulin) NOT (protein_name:insulin-like)

QUESTION 1.5:

1. How did you do this?
   by adding NOT (protein_name:receptor) to the query box.
2. How many hits are now left? How many of these are from Swiss-Prot?
   49 and 16

The contents of UniProt

QUESTION 2.1:

1. How many references are there in the insulin entry?
   37
2. Why do you think insulin is such a highly investigated protein?
   Because it is linked to a common and serious disease (diabetes) and used as a drug.

QUESTION 2.2:

1. Where do you find insulin?
   It is secreted from the cell (this is written just below the section heading. Under GO - Cellular component you can find additional locations mentioned, such as endoplasmic reticulum lumen, but these are temporary stages on the way to secretion).
2. Why do you think is it found there?
   Because it is a hormone - it has to travel through the bloodstream to influence other cells.

QUESTION 2.3: How long is the signal peptide and the propeptide, respectively?
24 and 31 amino acids.

Other databases linked from UniProt

QUESTION 2.4: How many mRNA sequences can you find links to, and what are their identifiers?
There are four mRNA sequences linked, and they are called: X70508, AY899304, BT006808, and BC005255.

QUESTION 2.5: What is the PDB Identifier of that structure?
The first structure to contain only one A and one B chain is 9R4E, number 3 in the list. Screenshot is here:

QUESTION 2.6: What is the InterPro identifier of the family that is just called "Insulin" and nothing else?
IPR004825.

Text format

QUESTION 2.7: Find the two feature table lines that describe the signal peptide and copy them to your answer.

FT   SIGNAL          1..24
FT                   /evidence="ECO:0000269|PubMed:14426955"

Advanced search

QUESTION 3.1: How many proteins did you find, and what was the search string (the text in the search field)?
12,497,853, of these 45,490 from Swiss-Prot
(ft_signal:*)

QUESTION 3.2: How many proteins do you find now, and what has the search string changed into?
3,922, they are all from Swiss-Prot
(ft_signal_exp:*)
Note that the "experimental" evidence is only found in Swiss-Prot entries, not in TrEMBL!

QUESTION 3.3: How many proteins do you find now, and what is the search string?
742
(ft_signal_exp:*) AND (organism_id:9606)

QUESTION 3.4: How many proteins are there in UniProt from Bacillus subtilis with the default TaxID [1423]?
903 results, of these only 63 from Swiss-Prot
(organism_id:1423)

QUESTION 3.5: How many proteins are there in UniProt from Bacillus subtilis in total (all strains and subspecies)?
5,341, of these 4,280 from Swiss-Prot
(taxonomy_id:1423)

QUESTION 3.6: How many proteins of maximum length 10 do you find?
9,009
(length:[1 TO 10])

QUESTION 3.7: How many proteins are now left?
1,324
(length:[1 TO 10]) AND (existence:1)

QUESTION 3.8: How many proteins are now left?
889
(length:[1 TO 10]) AND (existence:1) AND (fragment:false)

QUESTION 3.9: How many human non-fragment proteins of maximum length 10 do you find in UniProt?
6
(length:[1 TO 10]) AND (existence:1) AND (fragment:false) AND (organism_id:9606)

QUESTION 3.10:

Here they are in FASTA format:

>sp|P0DPR3|TRDD1_HUMAN T cell receptor delta diversity 1 OS=Homo sapiens OX=9606 GN=TRDD1 PE=1 SV=1
EI
>sp|P01858|TUFT_HUMAN Phagocytosis-stimulating peptide OS=Homo sapiens OX=9606 PE=1 SV=1
TKPR
>sp|P02729|GLUR_HUMAN Urine glycopeptide OS=Homo sapiens OX=9606 PE=1 SV=1
CEHSHDGA
>sp|P01358|GAJU_HUMAN Gastric juice peptide 1 OS=Homo sapiens OX=9606 PE=1 SV=1
LAAGKVEDSD
>sp|P02728|GLEM_HUMAN Erythrocyte membrane glycopeptide OS=Homo sapiens OX=9606 PE=1 SV=1
CEGHSHDHGA
>sp|P22103|PNEU_HUMAN Pneumadin OS=Homo sapiens OX=9606 PE=1 SV=1
AGEPKLDAGV

On your own, with and without AI

QUESTION 4.1: (taxonomy_id:562), 68,346 hits.

Using the question verbatim as a prompt yielded:

• Google: "Searching broadly for all E. coli strains captures a pool of 30 to 45 million entries" (!).
• Copilot: 68,344 UniProtKB protein entries (that's an error of 2).
• ChatGPT (free version): "The exact query to run in UniProt is: taxonomy_id:562" (Fair enough).

QUESTION 4.2: (taxonomy_id:83334), 5,999 hits.
Using the question verbatim as a prompt yielded:

• Google: "The raw total climbs into the hundreds of thousands of protein entries, primarily sitting unreviewed within TrEMBL" (!). However, asking it to provide the "UniProt query syntax" gave the correct search string.
• Copilot: 5,999 (correct). However, it falsely claims that the queries organism_id:83334 and taxonomy_id:83334 are equivalent. They are not (try for yourself).
• ChatGPT (free version): 9,080 (wrong). Asking it to provide the "UniProt query syntax" gave the wrong answer taxonomy_id:83334.

QUESTION 4.3: This is a question that does not have one single correct answer. A good attempt could be:
(protein_name:insulin) AND (gene:ins) NOT (protein_name:insulin-like) NOT (protein_name:"insulin related"), 602 hits

Using the prompt "How would you find insulin from as many organisms as possible from UniProt, without including entries that are not insulin?":

• Google: family:"insulin family" AND gene:ins AND length:[50 TO 150] which gives 661 hits (Not bad).
• Copilot:

((protein_name:"Insulin") OR
(protein_name:"Insulin-1") OR
(protein_name:"Insulin-2"))
NOT (
   protein_name:"Insulin receptor" OR
   protein_name:"Insulin-like growth factor" OR
   protein_name:"Insulin-induced"
)

which gives 10,353 hits with way too many false positives. But it also suggests family:"Insulin family" (8,929 hits) and combining that with filtering on protein name.

• ChatGPT (free version): does not come up with a useful UniProt query string, but suggests a lengthy workflow involving BLAST.

QUESTION 4.4:

• Our answer: (protein_name:"alpha globin") AND (taxonomy_id:9845) NOT (protein_name:"transcription factor"), 15 hits.
• Google: (gene:HBA* OR family:"hemoglobin alpha-type family") AND taxonomy_id:9895, 26 hits, actually better than our answer, because some entries were named "Hemoglobin subunit alpha" without "alpha globin" as an alternative name.
• Copilot:

taxonomy_id:9845 AND
(protein_name:"Hemoglobin subunit alpha" OR gene_exact:hba1 OR gene_exact:hba2 OR gene_exact:hba)

giving 37 hits without false positives, even better.

• ChatGPT (free version): required two prompts before it came up with the query string (protein_name:"hemoglobin subunit alpha" OR protein_name:"alpha-globin") AND taxonomy_id:9845, giving 58 hits including the "Alpha-globin transcription factor CP2" false positives.

QUESTION 4.5: (protein_name:"alpha-* globin") AND (taxonomy_id:8932), 2 hits.

• Google: gene:HBA* AND organism:"Columba livia" which does not work. However, it also suggests replacing the last part with taxonomy_id:8932, and then you get 3 hits. Whether the last TrEMBL hit ("Hemoglobin subunit alpha-D-like") is a false positive can be discussed.
• Copilot: gives several suggestion, of which only some are functional, e.g. organism_id:8932 AND (gene:hbaa OR gene:hbad OR protein_name:"alpha-A globin" OR protein_name:"alpha-D globin") (3 hits).
• ChatGPT (free version): organism_name:"Columba livia" AND (gene_exact:HBAA OR gene_exact:HBAD) (3 hits).

Note that you can shorten the last suggestion to organism_name:"Columba livia" AND (gene_exact:HBA*)