\end{center}
\begin{center}
-\begin{tabular}{cll}
+\begin{tabular}{clp{9cm}}
\hline
\multicolumn{1}{l}{\bf Field} & \multicolumn{1}{l}{\bf Type} & \multicolumn{1}{l}{\bf Description} \\\hline
{\tt DP} & {\tt uint16\_t[n]} & Read depth \\
{\tt GL} & {\tt float[n*G]} & Log10 likelihood of data; $G=\frac{A(A+1)}{2}$, $A=\#\{alleles\}$\\
-{\tt GT} & {\tt uint8\_t[n]} & {\tt haploid\char60\char60 7 | phased\char60\char60 6 | allele1\char60\char60 3 | allele2} \\
-{\tt \_GT} & {\tt uint8\_t+uint8\_t[n*P]} & {Generic GT; the first int equals the max ploidy $P$} \\
+{\tt GT} & {\tt uint8\_t[n]} & {\tt missing\char60\char60 7 | phased\char60\char60 6 | allele1\char60\char60 3 | allele2} \\
+{\tt \_GT} & {\tt uint8\_t+uint8\_t[n*P]} & {Generic GT; the first int equals the max ploidy $P$. If the highest bit is set,
+ the allele is not present (e.g. due to different ploidy between samples).} \\
{\tt GQ} & {\tt uint8\_t[n]} & {Genotype quality}\\
{\tt HQ} & {\tt uint8\_t[n*2]} & {Haplotype quality}\\
{\tt \_HQ} & {\tt uint8\_t+uint8\_t[n*P]} & {Generic HQ}\\
{\tt PL} & {\tt uint8\_t[n*G]} & {Phred-scaled likelihood of data}\\
{\tt PS} & {\tt uint32\_t[n]} & {Phase set}\\
%{\tt SP} & {\tt uint8\_t[n]} & {Strand bias P-value (bcftools only)}\\
-\emph{Integer} & {\tt int32\_t[n*X]} & {Fix-sized custom Integer; $X$ can be $1$, $A$ or $G$.}\\
-\emph{Numeric} & {\tt double[n*X]} & {Fix-sized custom Numeric; $X$ can be $1$, $A$ or $G$.}\\
+\emph{Integer} & {\tt int32\_t[n*X]} & {Fix-sized custom Integer; $X$ defined in the header}\\
+\emph{Numeric} & {\tt double[n*X]} & {Fix-sized custom Numeric}\\
\emph{String} & {\tt uint32\_t+char*} & {\tt NULL} padded concat. strings (int equals to the length) \\
\hline
\end{tabular}
BCF proposal).
\item Predefined {\tt FORMAT} fields can be missing from VCF headers, but custom {\tt FORMAT} fields
are required to be explicitly defined in the headers.
-\item A {\tt FORMAT} field with its name starting with `{\tt \_}' gives an alternative
- binary representation of the corresponding VCF field. The alternative representation
- is used when the default representation is unable to keep the genotype information,
- for example, when the ploidy is over 2 or there are more than 8 alleles.
+\item A {\tt FORMAT} field with its name starting with `{\tt \_}' is specific to BCF only.
+ It gives an alternative binary representation of the corresponding VCF field, in case
+ the default representation is unable to keep the genotype information,
+ for example, when the ploidy is not 2 or there are more than 8 alleles.
\end{itemize}
\end{document}