As most of the databases in the world, our database uses a special language named SQL (Structured Query Language) to execute queries.

Transcription

1 How to run SQL Queries against the database? As most of the databases in the world, our database uses a special language named SQL (Structured Query Language) to execute queries. This language allows different operations, but the most common is to retrieve the data. For this purpose SELECT statement is used. How SELECT works? Well, it is quite intuitive once you know the basic things. First it is good to know: The name of the table that you want to query. The structure of the table (fields, data types ) Let s start with an example: We want to retrieve all the data included in the table: all_info. The list of the tables available is shown in the drop down menu Table Info: Select the option "all_info", and the description of the table will appear: and scrolling:

2 This basically means that: The table all_info has 13 fields (columns) with the names: cluster_id unique cluster identifier (ranges from ) cluster_name ong name of the cluster (ra_c02_1 re assembly number 02 cluster 1) cluster_seq DNA sequence of the cluster cluster_length length of the sequence (in nucleotides) P_reads number of reads in POLYP stage S_reads number of reads in STROBILA stage E_reads number of reads in EPHYRA stage pep_seq predicted peptide sequence (or No Prediction tag) pep_length length of the peptide (in amino acids) score peptide prediction score (given by EstScan) start_nuc position of the first nucleotide of the ORF stop_nuc position of the last nucleotide of the ORF The table also indicates the type of the data: numeric (int or smallint) or text (varchar, mediumtext). Now we are going to construct and run our first query (retrieve all the data from a given table). Just type the following on the query field: SELECT * Which means: Retrieve (SELECT) all the fields (*) FROM table named: all_info.

3 NOTE: The syntax is case insensitive, so SELECT is the same as select or SElecT. The same happens with the column names and symbols. Press the button and a table with the results should appear (If not, check that you have written the query correctly): IMPORTANT: The results are limited to 1000 rows by default. Depending of the query, the results can be very large and memory consuming. Type "0" in the field for using no limit. Saving results as CSV In order to export the result to a file, the button can be used. This will save the results and the filters applied to the table in a CSV file that one can download to the computer and open in Ms Excel, for example.

4 Hiding Columns Question: How can I retrieve only data from columns cluster_id, P_reads, S_reads, E_reads and total_reads, for example? Answer: Just change the * for the field names separated by commas (,) except the last. Ordering Data Question: How do I order the data ascendant or descendant? Answer: There are two possibilities, but with a slightly difference: Click on the column s header. This will order the values of this column ascendant or descendant, but only affects to the values displayed. Here we sorted all the rows according to the expression level in strobila (see small arrowhead near "S_reads").

5 Use the clause ORDER BY in your query and use ASC or DESC to indicate the order. This affects to all the values in the database table. ORDER BY S_reads DESC

6 Filtering Data Question: How do I filter the data? Answer: There are two possibilities: Use the filter fields on the column header (text or numeric data). But, once time more, it only affects to the values displayed. If your result is greater than the limit specified (1000 by default) it could be some values on the results table that are not showed. For numeric values the following operators are available: equal to: = N greater than: > N lesser than: < N lesser or equal: <= N greater or equal: >= N range of values: N1.. N2

7 Use the clause WHERE and then the condition of your filter (it allows more complex searches). This is the recommended way: WHERE S_reads>1000 and E_reads<1000 Example with arithmetical and logical operations: WHERE (S_reads+1) / (P_reads+1) >= 500

8 WHERE (S_reads+1) / (P_reads+1) >= 500 and total_reads>2000 Adding an ORDER BY: WHERE (S_reads+1) / (P_reads+1) >= 500 and total_reads>2000 ORDER BY E_reads DESC

9 Selecting sub sets of genes Question: How can I retrieve the list of strobila specific genes? For example, we want to see only the clusters where more than 80% of all reads originate from the strobila stage. Moreover, we want to retrieve only the clusters where the total number of reads is >=10. Answer: We have to add additional conditions to our previous query. 80% threshold means that dividing the number of reads in strobila (S_reads) by the total number of reads (total_reads) we should get values >= 0.8. Total read number (total_reads value) should be >=10. So, we need to use a clause WHERE with two conditions: WHERE S_reads / total_reads >= 0.8 and total_reads >=10 As a result we will get a list of 345 strobila specific clusters (see Fuchs et al. Fig. 2B): And now let us get the list of the polyp specific genes. We need to change just one parameter (P_reads) and our new query will be: WHERE P_reads / total_reads >= 0.8 and total_reads >=10 As a result we will get a list of 336 polyp specific clusters (see Fuchs et al. Fig2B):

10 Getting the list of the ephyra specific genes is easy now. Here is the corresponding query: WHERE E_reads / total_reads >= 0.8 and total_reads >=10

11 Working with the microarray data The advantage of a relational database (like MySQL) is that it allows to work with large data sets and gives absolute flexibility in "asking" question of any level of complexity. It is easy to link different data types together, for example, sequence data with the corresponding expression values, peptide prediction, images and so on. For analysing the data one needs to describe the "question" as a set of mathematical and logical operators (in a similar way like in R, MatLab and similar programs). In the following examples we will use the table "array_normalized". This table contains mean signal values from independent experiments ((replicate_1+replicate_2+replicate_3)/3). Mean signal values across the stages (polyp, 14 days 5 Aza Cytidine, 14 days control,..., Ephyra) have been normalized based on the expression level of elongation factor 1 alpha (EF1α). Values in the table has not been subjected to logarithmic transformation. Log2 or Log10 transformation is important for presenting data in a form of a heat map, but for comparing expression that operation is not necessary. The table "array_normalized" contains 12 field (columns): id_entry unique entry identifier (primary key) id_oligo unique oligonucleotide name P_signal expression in POLYP (24h at 10 C) AZA14_signal expression in POLYP (14 days at 10 C, incubated in 5 Aza cytidine) CON14_signal expression in POLYP (14 days at 10 C, DMSO control) AZA16_signal expression in POLYP (16 days at 10 C, incubated in 5 Aza cytidine) CON16_signal expression in POLYP (16 days at 10 C, DMSO control) ES_signal expression in STROBILA with 1 segment LS_signal expression in STROBILA with 5 segments E_signal expression in EPHYRA (freshly detached) cl_name long name of the cluster (1 RA_1 cluster 1, rc_8 RA_8 cluster 8) cl_id unique cluster identifier (ranges from ) 1) To view all the values from the table type: SELECT * FROM array_normalized

12 The table with results should appear (if not, please check that the query has been correctly written): IMPORTANT: The results are limited to 1000 rows by default. Depending of the query, the results can be very large and memory consuming. Type "0" in the field for using no limit. 2) To find all the genes where expression in early strobila is 100 times stronger than in a polyp type:

13 SELECT * FROM array_normalized WHERE ES_signal / P_signal >= 100 ORDER by cl_id ASC IMPORTANT: Results will be sorted according to the cluster identifiers in ascending order (ORDER by cl_id ASC). You can also sort the results by clicking on the column's headers. 3) To identify genes which are up regulated during the temperature induction and might function as a strobilation inducer one will need a bit more complex query with many conditions (now we will describe the hypothetical model in Fig.4A): SELECT * FROM array_normalized where (P_signal+AZA14_signal+CON14_signal+AZA16_signal+CON16_signal+ES_signal+LS_signal+E _signal>=100) and P_signal<50 and CON14_signal>AZA14_signal and CON16_signal>AZA16_signal and ES_signal / P_signal>=5 and LS_signal / P_signal>=10 and LS_signal >= 1000 order by cl_id ASC

14 Here is the short explanation of the query: 1) We want to select genes which are expressed not extremely weak cumulative expression must be >= 100: (P_signal+AZA14_signal+CON14_signal+AZA16_signal+CON16_signal+ES_signal+LS_signal+E _signal>=100) 2) and the expression in the polyp stage must be weak: and P_signal<50 3) now we check that the genes are 5 AZA cytidin sensitive and the expression increases at cold temperature: and CON14_signal>AZA14_signal and CON16_signal>AZA16_signal and ES_signal / P_signal>=5 and LS_signal / P_signal>=10 4) expression in late strobila must be relatively high: and LS_signal >= ) ordering according to the cluster IDs (gene idenifiers): order by cl_id ASC As a result we get a list of potential strobilation inducers (27 clusters) represented by the heat map in Fig.4B in Fuchs et al. Expression dynamics of these genes follows the model represented in the Fig.4A. (See screenshot in the next page)

15 Thus, by using a simple set of commands one can extract a lot of information with nearly unlimited flexibility. It is also possible to combine data from several tables (JOIN statement). There is more information about databases and SQL at, for example: One can also consult MySQL Reference Manual at: