http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1998/37/i42/pdf/bi9809425.pdf Copied from acroread, removed the slashes and the p's Experiments from other papers were included here The Non-Two-State Sequences are not included. There is plenty of canonical pair data available without the two state data. (communication with Dave Mathews)

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2002/41/i40/pdf/bi026201s.pdf

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1997/36/i16/pdf/bi962608j.pdf The pdf file has tables as pictures, so they can't be copied. Took data from Zhi Lu's table, added second digit precision and STD

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1998/37/i04/pdf/bi972050v.pdf The pdf file has tables as pictures, so they can't be copied. Took data from Zhi Lu's table, added second digit precision and STD

These are considered hairpin loop terminal mismatches in the paper. I don't know the ERR-dG-37, I just put 5%.

http://www.rnajournal.org/cgi/reprint/6/4/608.pdf Did not include Table 4B (communication with Dave Mathews).

These are considered hairpin loop terminal mismatches in the paper. It's not clear what is the experimental error for the duplexes. 5% sounds reasonable (communication with Dave Mathews).

http://nar.oxfordjournals.org/cgi/reprint/20/4/819 STD-dG-37 0.3 sounds more conservative than 0.2 (communication with Dave Mathews).

http://nar.oxfordjournals.org/cgi/reprint/19/21/5901 Some of the data in Table 1 are the same as Table 1 from Antao_Tinoco_1992, however the values are different. I included them, and the parser should be aware that there can be duplicates with which it should take the average (communication with Dave Mathews). In fact, I'm doing regression, so I don't think the parser needs to know anything.

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1994/33/i47/pdf/bi00251a042.pdf I've also included Table 3 as motif stem (communication with Dave Mathews).

These are considered hairpin loop terminal mismatches in the paper. By convention, we use the results from 1/Tm vs log Ct/4 plots, and not the ones from average of curve fits (communication with Dave Mathews).

The pdf file has tables as pictures, so they can't be copied. Took data from Zhi Lu's table, added second digit precision and STD

http://nar.oxfordjournals.org/cgi/reprint/21/16/3845 The pdf file has tables as pictures, so they can't be copied. Took data from Zhi Lu's table, added second digit precision and STD It looks like having triphosphate at one end gives different values

http://nar.oxfordjournals.org/cgi/reprint/16/24/11725 Took data from Zhi Lu's table, added second digit precision and STD Whenever there are 2 dangling ends on the same side, they can be considered to be terminal mismacthes.

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1986/25/i11/pdf/bi00359a019.pdf No dG values, I computed them from enthalpy and entropy (communication with Dave Mathews). The following are self-complementary (self complementary = a strand can form a duplex with another identical strand - communication with Dave Mathews).

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1989/28/i02/pdf/bi00428a049.pdf Structures written by hand, values copied from acroread Whenever there are 2 dangling ends on the same side, they can be considered to be terminal mismacthes. Error 5% sounds reasonable (communication with Dave Mathews).

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2002/41/i33/pdf/bi025781q.pdf If the bulge can be in more than one place, I added one line per place I put error of 5% on dG-37.

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1990/29/i01/pdf/bi00453a038.pdf

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1991/30/i46/pdf/bi00110a015.pdf All the G-U were considered pairs, except (GGUC)2, which was considered 2x2 internal loop, according to Mathews_Turner_1999. (This may give non-full rank matrix) I did not include the non-two-state transitions. They are not good (communication with Dave Mathews).

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2001/40/i08/pdf/bi002564l.pdf Self-complementary means that a strand can form a duplex with another identical strand. Note that this does not require the symmetry penalty unless the structure is symmetric. The following is self-complementary and symmetric (communication with Dave Mathews).

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/2000/39/i31/pdf/bi000229r.pdf The Table is as picture :(, I had to write all the data by hand. The following structure has A-A pair at the end. I think there is a mistake, considering Figure 1. The second one must be a U. the following structure has A-A pair at the end. I think there is a mistake, considering Figure 1. The second one must be a U. The following two are outliers: the error is greater than 1 kcal/mol

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/2000/39/i31/pdf/bi000229r.pdf The Table is as picture :(, I had to write all the data by hand. for the following, we don't have the value for the STD-dG-37. I just put 0.2

http://pubs.acs.org/cgi-bin/sample.cgi/bichaw/2003/42/i48/pdf/bi0301587.pdf Since The following 3x3, 2x3 etc parameters are not included in my model yet, A lot of the following lines give conflicts. I removed all of them for now.

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1991/30/i26/pdf/bi00240a013.pdf I don't need the values in parentheses (i.e. parameters corrected for single-strand extinction coefficients). I used the out of parantheses values, to be consistent to the rest of the database (communication with Dave Mathews).

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1996/35/i50/pdf/bi961789m.pdf The Table is as picture :(, I had to write all the data by hand. the following structure has U-U pair at the end. I think there is a mistake, the first one must be an A.

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1995/34/i10/pdf/bi00010a009.pdf

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1994/33/i37/pdf/bi00203a033.pdf

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1991/30/i33/pdf/bi00247a021.pdf

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1997/36/i41/pdf/bi971069v.pdf All the adjacent G-U were considered pairs, except (GGUC)2, which is considered 2x2 internal loop, according to Mathews_Turner_1999. If only one of the 2 pairs is G-U, and the other is a mismatch (e.g. A-G), the G-U was not considered paired, and the A-G 1x1 mismatch, but all was a 2x2 internal loop.

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1999/38/i43/pdf/bi991186l.pdf Did not use the non-two-state melting values. This contains a G-U "internal loop", but we consider it to be a base pair.

http://www.dnasoftware.com/Science/Publications/pdf/ja00011a039%20JSL%20JACS1991.pdf I didn't include the duplexes with modified nucleotides (I, P, 2), because the nearest neighbour model doesn't deal with modified nucleotides. The following two are the same and have different values, because they are from different sources. I included both, and the regression will take care of them.

http://www.pnas.org/cgi/reprint/101/19/7287 Written by hand, from the supporting text

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1996/35/i45/pdf/bi9615710.pdf All the G-U were considered pairs, except (GGUC)2, which is considered 2x2 internal loop, because they are non-nearest-neighbor, according to Mathews_Turner_1999. (communication with Dave Mathews)

http://nar.oxfordjournals.org/cgi/reprint/27/4/1118 Written by hand, couldn't copy from acroread I considered the U-G mismatches to be base pairs, and not 1x1 internal loops (communication with Dave Mathews).

http://nar.oxfordjournals.org/cgi/reprint/23/2/302 The following are self-complementary. In general, there is no reason why these need to be symmetric, so I would leave the symmetry penalty off. (communication with Dave Mathews).

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1997/36/i49/pdf/bi9716783.pdf I tried to include Table 2 and 3, but it's not clear to me what those numbers represent exactly (the difference is between what and what?). So for now, they are not included. I skipped this data and only used optical melting data (communication with Dave Mathews). http://pubs.acs.org/cgi-bin/searchRedirect.cgi/bichaw/2004/43/i40/pdf/bi049168d.pdf

http://pubs.acs.org/cgi-bin/searchRedirect.cgi/bichaw/2005/44/i08/pdf/bi048079y.pdf At the beginning of every Seq2, there was a P for purine. Following communication with Dave Mathews and Brent Znosko, I left it as P, and added a new parameter in my model to account for this particular dangling end. Brent Znosko says: "Purine is a modified nucleotide, essentially adenosine without the exocyclic amino group. This is sometimes used to investigate the role of the exocyclic amino group in stabilization; an adenosine would be replaced by a purine. [...] Purine was used as a dangling end. This was used to give added stability to the duplex. I'm assuming that Gang Chen used purine to distinguish NMR peaks from this nucleotide from the adenosines in the 3x3 loop. Since it is used as a dangling end here, my suggestion to you would be to treat this as an adenosine (as we did in the paper as well). My guess is that there is not much difference between a dangling purine and a dangling adenosine, since the amino group in a dangling adenosine probably does not play a significant role in the stabilization of the duplex.". Dave Mathews says: "I agree with Brent's logic. There is, however, a duplex in the study that is melted with the dangling purine and no other unknown parameters, so I solved for the stability of the dangling purine. I calculated -1.1 for the purine in this context and a dangling A this context is worth -1.7 kcal/mol. How about subtracting -1.1 from all the duplexes in this case and removing the dangling end? A more rigorous approach might be to add a new parameter to your list of parameters for the dangling purine and let that get fit (because the calculated -1.1 is subject to experimental error)".

http://pubs.acs.org/cgi-bin/searchRedirect.cgi/bichaw/2005/44/i50/pdf/bi051659q.pdf The sequence-pattern for the following sequences is different from the pattern of the above. Mistake corrected on Apr 21, 2008. The values for the following three duplexes are obtained from the best fit of calculated transitions. I included them, but I have to assign them higher errors. Also, the minimum error in general should be 0.1 (communication with Dave Mathews).

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1986/25/i19/pdf/bi00367a061.pdf dG values are not given. I did not use the temperature-dependent numbers. I calculated DG from DH and DS. The temperature dependent numbers were generated using an alternative method for data analysis. (communication with Dave Mathews) I removed the predicted values, i.e. the ones in parantheses, and with superscript f.

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2001/40/i23/pdf/bi0029548.pdf Here, state 0 is not the completely unfolded state, but it's the hairpin formed by sequence0.

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2002/41/i03/pdf/bi011441d.pdf Here, state 0 is not the completely unfolded state, but it's the hairpin formed by sequence0.

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2007/46/i46/pdf/bi701311c.pdf Only included the 24 duplexes without footnote, because some footnoted duplexes are from other papers that are included here, or are not two-state melts or other reasons. The following had a G-G non-canonical base pair in the paper, but it was a typo, the first G is actually a C (communication with Brent Znosko).

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2007/46/i15/pdf/bi0616910.pdf I didn't include this paper, because it contains experiments with Inosine, and the results are different from the case when I is replaces by A or G. http://pubs.acs.org/cgi-bin/asap.cgi/bichaw/asap/pdf/bi701024w.pdf There is a typo in the original paper, mismatch C-U should be base pair C-G (communication with Brent Znosko).

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2004/43/i50/pdf/bi049256y.pdf The first entry of Table 1 has a P (purine) at the 3' end of the second strand. I left it P, like in Chen_Turner_2005. Entries 2 and 3 contain I, so I skipped them. Entries 4 and 5 contain P in the middle, I skipped them. (communication with Dave Mathews).

http://www.rnajournal.org/cgi/reprint/11/4/512

I don't know the ERR-dG-37, I just put 5%.

http://nar.oxfordjournals.org/cgi/reprint/34/11/3338 Also included duplexes from other references, because they weren't included elsewhere in this file. The duplexes from other papers first. Only "average of curve fits" data are reported, but I thought I would better use these then none. Moved the ones with only one dangling base to the end (the ones marked with f).

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2004/43/i37/pdf/bi049954i.pdf

I think this table means to consider dangling free bases as mismatches in hairpin loops, just like in Dale_Serra_2000.

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2006/45/i05/pdf/bi051750u.pdf I think in Tables 1-5, the "L" subscript of dG37 should be for the last column. If I'm wrong, then the dG values below are wrong. Dave Mathews agrees.

I think this table means to consider dangling free bases as mismatches in hairpin loops, just like in Dale_Serra_2000.

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1987/26/i14/pdf/bi00388a058.pdf The values were measured spectrophotometrically. Optical melting absorbances are measured with a spectrophotometer, so spectrophotometrically is another way of stating by optical melting, although not as specific. The free energy changes will be calculated from enthalpy and entropy.

http://nar.oxfordjournals.org/cgi/reprint/32/5/1824

http://pubs.acs.org/cgi-bin/sample.cgi/bichaw/2006/45/i39/pdf/bi0605787.pdf

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2007/46/i44/pdf/bi700802s.pdf No optical melting values in this paper. http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2007/46/i06/pdf/bi061350m.pdf

http://nar.oxfordjournals.org/cgi/reprint/34/13/3609 I don't need this paper, because it contains 2'-O-methyl RNA, and experiments are done at 0.1M NaCl instead of at 1M NaCl. A 2'-O-methyl is a modification to the sugar. Instead of a 2'-OH, there is a 2'-O-methyl = 2'-O-CH3 group. Therefore, these sequences are not natural and should be excluded (communication with Dave Mathews). http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2006/45/i22/pdf/bi0524464.pdf I'm not sure if I can get anything out of this paper: all new experiments have D,M,P,I in the sequences. (Dave Mathews agrees). http://pubs.acs.org/cgi-bin/sample.cgi/bichaw/2006/45/i12/pdf/bi052060t.pdf Cannot be copied. Data in text format obtained from Gang Chen. The P's (purine) in this paper is dangling. Treated the same as in Chen_Turner_2005.

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/1999/38/i46/pdf/bi991774z.pdf I didn't include this data, because it's not done at 1M NaCl, as the other ones. http://pubs.acs.org/cgi-bin/article.cgi/bichaw/1997/36/i26/pdf/bi970122c.pdf

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/1996/35/i42/pdf/bi961310q.pdf

http://pubs.acs.org/cgi-bin/article.cgi/bichaw/2008/47/i14/pdf/bi7020876.pdf

http://pubs.acs.org/cgi-bin/archive.cgi/bichaw/1990/29/i37/pdf/bi00489a044.pdf Added on Apr 21, 2008, for version 1.3

http://www.nature.com/nature/journal/v331/n6153/pdf/331283a0.pdf

Table 1 is for T4-32 not T-35 (the order of T-32 and T-35 is swapped as compared to Figure 1). Until April 11, 2009, there was a mistake: the sequence and structure for T-32 had the experimental free energy of T-35 and vice-versa. This mistake was fixed on April 11, 2009.

I couldn't find this paper online