Candidate Oligos for every Stanford Oligo Chip track

Oligos were chosen for every Sanger22 annotation on chr22 as well as about 2000 other genes. Two oligos were chosen with a 3' bias, two with a 5' bias, and two with no bias. For this purpose exons are defined to include 3' and 5' UTRs.

The strategy

These oligo selections are based on the following ideas:

• Oligos should have minimum secondary structure as they must be available for hybridization.
• Oligos should be unique in genome if possible. No repeats, should not Blat or Blast other places in genome.
• If using oligo-dT for RT-Priming oligos should be in 3' end of gene transcript (including UTR).
• Oligos should have a uniform hybridization temperature if possible. All oligos must be hybridized at same temperature, want to minimize cross hybe yet maximize signal.

Currently we don't have data to identify which parameters are more important than others. Also, some of these scores are overlapping (i.e. if tm is limited then high secondary structure is less likely). See below for histograms of these criteria.

The Details:

The Algorithm

• Step through each exon at a step size proportional to the size of the exon examining possible oligos, excluding areas that are RepeatMasked.
• Score each oligos for: Tm difference, distance from 3' end, secondary structure, and an Affymetrix heuristic.
• Look through candidate probes remembering the maximum score for each score.
• Each score is then normalized by dividing by the maximum and then the normalized scores are combined as an average and oligos are sorted to find the best overall score.
• Oligos with the best combined normalized scores are blatted until one is found that has a blat score below a given threshold.
• As oligos are chosen, candidate oligos that overlap those already chosen are discarded.
• If no scores pass the blat score or not enough oligos have been chosen just pick oligos that have the best combined score.

About the scores:

• Tm: Formulas for calculating Tm taken from: "A unified view of polymer, dumbbell, and oligonucleotide DNA nearest-neighbor thermodynamics" John SantaLucia, Jr. PNAS, Vol 95, pp 1460-1465 February 1998.
A web version called Hyther exists.
• Secondary Structure: Calculates the Gibbs free energy of the best secondary structure using libraries from the RNAstructure program.
• Affy Heuristic: 1 if oligo passes heuristics derived from that published by Affymetrix "Nature Biotechnology" vol. 14, Dec, '96) are satisfied, 0 otherwise. The heuristic is as follows:

     no more than 9 A's in window of 20 
     no more than 9 T's in window of 20
     no more than 8 C's in window of 20
     no more than 8 G's in window of 20
    
     no more than 6 A's in window of  8
     no more than 6 T's in window of  8
     no more than 5 C's in window of  8 
     no more than 5 G's in window of  8
  

• 3' Dist: Distance from end of oligo to 3' end of target sequence.
• Blat Score: Blat score of second most homologous region in the genome. If no inserts this is approximately the number of base pairs that match.

Histograms of Scores

Histograms are from the Stanford picked gene set.

Secondary structure measured in Gibb's Free energy, higher scores are better.	Blat (similar to blast) histogram, lower scores are better.
Melting temperatures, scores over 100C do happen in algorithm.	Percentage GC, not used in algorithm but presented anyway.

Please note that all coordinates are relative to the '+' strand while all oligo sequences are 5'->3'. This means that all sequences displayed are part of the sense strand. So if the oligo is represented in the database as being on the '-' strand and starts at 1 and ends at 5 of 'atgcatgc' the '+' sequence of the probe would be 'tgcat' but that is 3'->5' on the '-' strand so the sequence in the sequence would be the reverse complement 'atgct'.