Similarity Maps for Antipsychotic Medications - Dopamine and Serotonin Receptor Affinities : Part Three
Introduction
In Part Two we derived a similarity map for 12 medications' dopamine receptor affinity profile. In this post, we move on to compare serotonin and dopamine receptors and their similarity maps.
Additional Data and Preprocessing
Before continuing, some changes in plans with regards the medications we'll use. Paul Morrison and colleagues have a forthcoming book Advanced Prescribing in Psychosis. Paul was kind enough to give me a draft to read. In the book's tables, they've included some medications which I hadn't been able to find in the the PDSP database [1]. So, I scraped some papers to add in Ki values for brexipiprazole [2], asenapine [3] and lurasidone [4].
The result was some sparsity in the tables of medication/receptor pKi values because of receptor synonyms and missing data (e.g. some older medications don't have pKi values for some isoforms of 5HT and DA receptors). We use the following information to combine the PDSP records with using more contemporary nomenclature:
- D2 is often divided into D2 Long and D2 Short which are synonymous with D2A and D2B; see [5] and [6]. So, we combine D2 Long with D2A, and D2 Short with D2B.
- D1A is synonymous with D1 - see [7] for details - so again, we combine them
- 5HT-7A and 5HT-7B are synonymous with the 5HT-7 long and short isoforms respectively [8], so they get combined also.
Next, we find that the D2 isoforms are often profiled in newer, but not older medications (i.e. older medications tended to have D2 and newer D2 Long). Inspecting the data, there are only minor differences (easily within an order or magnitude) where D2 and D2 Long or Short are reported, so we take the means and arrive at an aggregated D2 affinity. For D4, newer medications are sub-typed on D4.4 and D4.2, with older medications only having D4. Again, aggregating solves the problem of differences in newer and older medications and all where within an order of magnitude, so we can arrive at an overall D4 affinity.
The second revision is to 'quantise' the previous heatmap representation; in discussions with colleagues, it was suggested that the real-valued pKi values don't add much and really, users of these tables want "High", "Moderate", "Low" banding, and typically in prescribing guidelines and textbooks, these are denoted as e.g. "+++", "++" and "+". We follow the convention used in Morrison et al's forthcoming book:
| Affinity | Ki (nM) |
pKi -log( Ki ) |
|---|---|---|
| Insignificant | > 1000 | < 6 |
| Low | 100 - 1000 | 6 - 7 |
| Moderate | 10 - 100 | 7 - 8 |
| High | 1 - 10 | 8 - 9 |
| Very High | < 1 | > 9 |
Note, the logarithm is base 10.
Heatmaps
Here's the resulting quantized heatmaps (DA receptors on the left, 5HT right) corresponding to the textbook's tables - they can be enlarged by clicking.
Notice that, especially for serotonin receptors, there's still quite a bit of sparsity. To build similarity maps, we need to filter on only the medications and receptor combinations with complete data. This results in some loss - for example, when we look at 5HT similarities, we lose flupenthixol because we have no practically no 5HT receptor information). The balance is to maximize medications included simultaneously with maintaining complete receptor coverage.
Similarity Maps : Dopamine DA1-DA4
For dopamine, we can extract complete data for DA1 through DA4 (but DA5 was too sparse to use). Another change I've made is to drop colouring by generation, instead, replacing it by the total (sum) of affinities over DA1-DA4; so for example, a medication with overall strong or weak affinity for all DA receptors is coloured dark red and green respectively.
The explanation for how these maps (and the connecting arrows) are derived via MDS can be found in the previous post. As before, a directed arrow from A to B is read "For medication A, the most similar medication is B" and the relative positions on the plane reflect similarities (but with some distortion introduced by MDS)
Figure One : AMI = AMISULPRIDE; ARI = ARIPIPRAZOLE; ASE = ASENAPINE; BRE = BREXPIPRAZOLE; CHL = CHLORPROMAZINE; CLO = CLOZAPINE; FLU = FLUPENTHIXOL; HAL = HALOPERIDOL; ILO = ILOPERIDONE; LOX = LOXAPINE; LUR = LURASIDONE; MEL = MELPERONE; OLA = OLANZAPINE; PER = PERPHENAZINE; PIM = PIMOZIDE; QUE = QUETIAPINE; RIS = RISPERIDONE; SER = SERTINDOLE; SUL = SULPIRIDE; TRI = TRIFLUOPERAZINE; ZIP = ZIPRASIDONE
Similarity Maps : 5HT-1A, 2A, 2C, 6 and 7
For the same group of medications, but for serotonergic receptors. The sparsity of coverage means that only 5HT1A, 2A, 2C, 6 and 7 are available for all the medications.
Figure Two: AMI = AMISULPRIDE; ARI = ARIPIPRAZOLE; ASE = ASENAPINE; BRE = BREXPIPRAZOLE; CHL = CHLORPROMAZINE; CLO = CLOZAPINE; HAL = HALOPERIDOL; ILO = ILOPERIDONE; LOX = LOXAPINE; MEL = MELPERONE; OLA = OLANZAPINE; PER = PERPHENAZINE; PIM = PIMOZIDE; QUE = QUETIAPINE; RIS = RISPERIDONE; SER = SERTINDOLE; TRI = TRIFLUOPERAZINE; ZIP = ZIPRASIDONE
Next Steps ...
Up till now, we've seen similarity maps where - for DA and 5HT - we've compressed a 4 and 5 dimensional space (of affinity profiles) down to a 2D embedding. The 'axes' are systematically related to distances (because of how MDS algorithms work), but not in a way that has a clear meaning, and we've treated them as 'unitless' dimensions. In the next post, I'll explore ways of visualizing both DA and 5HT simultaneously, which will require some experiments compressing the 4 and 5-D spaces down to a single dimension each, and then plotting DA and 5HT on orthogonal axes.
References:
- The Multiplicity of Serotonin Receptors: Uselessly diverse molecules or an embarrasment of riches? BL Roth, WK Kroeze, S Patel and E Lopez: The Neuroscientist, 6:252-262, 2000
Maeda, K., Sugino, H., Akazawa, H., Amada, N., Shimada, J., Futamura, T., ... & Pehrson, A. L. (2014). Brexpiprazole I: in vitro and in vivo characterization of a novel serotonin-dopamine activity modulator. Journal of Pharmacology and Experimental Therapeutics, 350(3), 589-604.
Shahid, M., Walker, G. B., Zorn, S. H., & Wong, E. H. F. (2009). Asenapine: a novel psychopharmacologic agent with a unique human receptor signature. Journal of psychopharmacology, 23(1), 65-73.
Ishibashi T, et al (2010). Pharmacological profile of lurasidone, a novel antipsychotic agent with potent 5-hydroxytryptamine 7 (5-HT7) and 5-HT1A receptor activity". J. Pharmacol. Exp. Ther. 334 (1): 171–81.
Harding SD, et al. (2018) The IUPHAR/BPS Guide to PHARMACOLOGY in 2018: updates and expansion to encompass the new guide to IMMUNOPHARMACOLOGY. Nucl. Acids Res. 46 (Issue D1): D1091-D1106. doi: 10.1093/nar/gkx1121http: Full text here
Hayes, G et al, 1992 Structural subtypes of the dopamine D2 receptor are functionally distinct: expression of the cloned D2A and D2B subtypes in a heterologous cell line., Molecular Endocrinology, Volume 6, Issue 6, 1 June, Pages 920–926, https://doi.org/10.1210/mend.6.6.1323056
Huntley RP, et al (2014) The GOA database: Gene Ontology annotation updates for 2015. Nucleic Acids Research 2014 doi: 10.1093/nar/gku1113. https://www.ebi.ac.uk/QuickGO/term/GO:0031748
Heidmann, D. E., et al (1997), Four 5‐Hydroxytryptamine7 (5‐HT7) Receptor Isoforms in Human and Rat Produced by Alternative Splicing: Species Differences Due to Altered Intron‐Exon Organization. Journal of Neurochemistry, 68: 1372-1381. doi:10.1046/j.1471-4159.1997.68041372.x
Wishart DS, Feunang YD, Guo AC, Lo EJ, Marcu A, Grant JR, Sajed T, Johnson D, Li C, Sayeeda Z, Assempour N, Iynkkaran I, Liu Y, Maciejewski A, Gale N, Wilson A, Chin L, Cummings R, Le D, Pon A, Knox C, Wilson M. DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic Acids Res. 2017 Nov 8. doi: 10.1093/nar/gkx1037.
