隠遁Chemistと愛(AI)の交換日記
2006年1月1日に無事リリースされた。たまにクラッシュする。細かくバージョンアップして行く必要はありそうだ。
結局、距離の式の計算でEuclidタイプの計算式は導入されなかった。
HSP Distance (Euclid) =sqrt(4*(δDs-δDl)2 +(δPs-δPl)2 +(δHAcs-δHAcl)2 +(δHbss-δHbsl)2 )
酸塩基を導入した時に使えるのはBeerbowerタイプの式のみなる。
HSP Distance (Beerbower) =sqrt(4*(δDs-δDl)2 +(δPs-δPl)2 +2*(δHAcs-δHAcl)2 *(δHBss-δHBsl) )
Beerbowerタイプの式を使うと、アルコールに溶けないような非極性の溶質(s)を使った場合、クラッシックの3D-HSPを使った場合より結果が悪くなる。
それは、溶質のδHAcsが0.1で溶媒のδHAclもほとんどの溶媒(l)で0.1なので、水素結合項の寄与が０になってしまうからだ。
Abbott先生にいくら説明しても、自分にはEuclid式は必要ないの一点張りなので諦めて欲しい。
AIよ。HSPiP ver.6.2、山本分はチェック終了。軽微な修正でOKが出た。のブログに書いたが、HSPiPを騙して、3次元で計算すればEuclidもどきで計算できる。
正しくは、Pirika pro Add-Onを使うしかない。
日本語マニュアルに反映するのは少し先になるだろう。（購入者には先に6.1のマニュアルを渡し、後日追加分を送る事にする）
Abbott先生のリリースノート
V6.2.02 January 2026
Thanks to the kind permission of Prof Yoshida, Japan’s leading Polymer AI expert, we have included his PolyOmics dataset of key properties of polymers found via their Polymer SMILES. When you click the Help button on the PolyOmics tab in DIY you will find links to both the key paper and the key datasource provided by Prof Yoshida and his team. For computational reasons, they calculate δD, δTot and δE where E is an amalgam of δP and δΗ.
Ctrl-F on the main form takes you to the text box ready to enter your search text
Some molecules with many @ symbols in their SMILES were rejected as being too long in DIY. That has been fixed
A new button on the main form allows you to open new HSPiP for comparisons
Sphere Explorer allows you to create your own mono- or multi-sphere fits, manually or automatically. If you use Polymers88A and B (PEMA and PMMA) as examples you can explore the idea that different parts of the polymer might interact differently when near borderline solubility.
Behind the scenes the code has been cleaned up, reducing the number of supporting .dll files and making it easier to add new features into the future. Calculations are now generally faster.
While doing so, inconsistencies in how we show the number of decimal places have been sorted out, making tables look much cleaner. For example, reporting an estimated HSP to 2 decimal places or a BPt or MPt to one decimal place makes no sense, so HSP are 1 decimal place and BPt and MPt are integers.
The datasets now include δHD-L and δHA-L, the largest possible H-bond interactions if there is a chance to overcome “mean field” behaviour.
In the Solubility tab of Y-MB, the solubility data are now graphed.
The QSAR has been updated with new cleaner tables, minimum and maximum charges have been added along with the extra δHD-L and δHA-L columns. There is also the option to include RDKit molecular informatics parameters if you find them helpful for creating a good QSAR fit.
We now use an updated version of RDKit and have added access to some of the molecular informatics parameters it can provide. You can do this from a tab in DIY and also within the QSAR.
There are many ways the same molecule can be described in SMILES. Ethanol can be OCC or CCO. This makes it hard to search for a SMILES. Thanks to RDKit, everything has been converted to unique Canonical SMILES. If, for example, you look for OCC, it is first converted to the canonical CCO and found in the dataset.
For years we have used X to signify where monomer units join for Polymer SMILES. So polyethylene oxide is XCCOX. There is now a more modern standard of *, so PEO is *CCO*. You can now use either, though internally we stick with X for backwards compatibility.